E Book Ta PDF
E Book Ta PDF
E Book Ta PDF
Technical Assistance
G.M. Pathak
Neelam Chandra
Abha Kant
Designed by
Rakesh Kr Saini
1. Introduction
India is primarily an agriculture-based country and its economy largely depends upon
agriculture. Presently, contribution of agriculture about one third of the national GDP and
provides employment to over seventy percent of Indian population in agriculture and allied
activities. Therefore, our country’s development largely depends upon the development of
agriculture. The agricultural production information is very important for planning and
allocation of resources to different sectors of agriculture. Agricultural statistics in India have
a long tradition. Artha Shastra of Kautilya makes a mention of their collection as a part of the
administrative system. During the Moghul period also some basic agricultural statistics were
collected to meet the needs of revenue administration. Ain-e-Akbari is most important
document that throws great light on the manner in which statistics were collected during the
moghul period. After the Moghul period British rule started Ryotwari System, introduced
during 18th Century to collect land revenue. In 1866, the British Government initiated
collection of agricultural statistics mainly as a by product of revenue administration and these
reflected the then primary interest of the Government in the collection of land revenue.
Subsequently, the emphasis shifted to crop forecasts designed primarily to serve the British
trade interests. On a representation made by a leading firm of Liverpool, trading in wheat,
the preparation of wheat forecast was taken up in 1884 and the land utilization statistics are
available in the country since 1884. By 1900, oilseeds, rice, cotton, jute indigo and sugarcane
had also been added to the list of forecast crops.
The system of agricultural statistics generates valuable statistics on a vast number of
parameters. Some of the very important statistics are land-use statistics and area under
principal crops through the Timely Reporting Scheme (TRS) and also on complete
enumeration basis, yield estimates through the General Crop Estimation Surveys (GCES), the
scheme of Cost of Cultivation Studies (CCS), checks the reliability of TRS and yield
estimates through the scheme of Improvement of Crop Statistics (ICS) , cost of production
estimates, agricultural wages, irrigation statistics, conduct of Agricultural Census and
Livestock Census on quinquennial basis. The system of agricultural statistics also generates
data on livestock products through the scheme of Integrated Sample Survey (ISS), collects
wholesale and retail prices, conducts market intelligence and observes rainfall and weather
conditions.
2. Crop Area Statistics
The country can be divided into four broad categories with respect to collection of area
statistics namely (i) Temporarily settled states, (ii) Permanently settled states and (iii) Other
regions.
(i) Temporarily Settled States
The system of temporarily settlements was introduced in our country in 1892, with a view to
fix land revenue for a period, which was subject to change at the time of the next settlement.
Ordinarily, the interval between two settlements was 25 to 30 years. In order to determine the
land revenue and to make estimates of production forecast detail statistics are to be collected
about land revenue, land value etc. In temporarily settled areas the information on crop area
statistics are collected by the village accountant or Patwari and are recorded in a register
which is popularly known in northern India as Khasra. The village accountant has been called
by different names in different parts of the country such as Karnam in South, Telatti in
Maharashtra, Karamchari in Bihar, lekhpal in Uttar Pradesh etc. This category covers around
86% of total reported area of the country.
The crop area statistics collected by village accountant are on the basis of complete
enumeration called girdawari. The village accountant is to visit each and every field of the
village in each crop season and record the information such as area under different crops/land
use categories and its status in standard forms called Khasra register. The work of village
accountant is supervised by immediate superior officer known by the name of Quanungo in
northern India. Most of the geographical areas of temporarily settled states are cadastrally
surveyed and detailed maps are available in tehsil and district Headquarters. The statistics
obtained by different village accountants are aggregated to get the crop area statistics at
higher administrative units such as blocks, tehsils, district, states etc. This system of data
collection is being followed in 18 states namely Andhra Pradesh, Assam (excluding hill
districts), Bihar, Chattisgarh, Goa, Gujarat, Haryana, Himachal Pradesh, Jammu & Kashmir,
Jharkhand, Karnataka, Madhya Pradesh, Maharashtra, Punjab, Rajasthan, Tamil Nadu,
Uttaranchal and Uttar Pradesh and 5 union territories Chandigarh, Dadra and Nagar Haveli,
Daman & Due, Delhi and Pondichery.
(ii) Permanently Settled States
There are three states namely Kerala, Orissa and West Bengal which come under category of
permanently settled states. In case of these states land revenue was permanently fixed and
question of revision ordinarily did not arise. In these states there is no system of recording
details of area statistics as there is no permanent revenue staff for a village like village
accountant as in the case of temporarily settled area. Initially there was no uniform system of
collecting area statistics in these regions. The police Chaukidar or village headman was
usually providing the statistics on the basis of guess work which were quite unreliable. In
order to improve the quality of these statistics in the permanently settled states, presently, the
area statistics in these states are collected by specially appointed field staff under the scheme
known as “Establishment of Agency for Reporting Agricultural Statistics (EARAS)” which
was initiated in 1968-69. In the States covered by EARAS, the complete enumeration of all
fields (survey numbers) i.e. girdawari is conducted every year in a random sample of 20%
villages of the States, which are selected in such a way that during a period of 5 years, the
entire state is covered. This category covers around 9% of total reported area of the country.
(iii) Other Regions
The remaining eight states in North Eastern regions namely Arunachal Pradesh, Manipur,
Meghalaya, Mizoram, Nagaland, Sikkim and Tripura and two other union territories namely
Andaman and Nicobar Islands and Lakshdweep do not have a proper reporting system,
though states of Tripura and Skkim (except some minor pockets) are cadastrally surveyed.
In these regions compilation of area statistics are based on conventional methods in which
estimates are reported by village choukidars on the basis of personal assessment. This
category covers around 5% of total reported area of the country.
3. Timely Reporting Scheme (TRS)
In order to reduce the time-lag between the sowing and availability of estimates of area and
harvesting of crops and availability of estimates of production, a Centrally sponsored scheme
for Timely Reporting of Estimates of Area and Production of Principal Crops (TRS), was
initiated by the Ministry of Agriculture and Irrigation in the year 1968-70. The basic
objective of TRS is to reduce time lag for making available area statistics of major crops in
I.2
addition to providing the sample frame of selection of crop growing fields for crop cutting
experiments in permanently settled states. Under the scheme, the villages in each stratum
(tehsil/revenue inspector circle/patwari circle etc.) are divided into 5 independent non-
overlapping sets, each comprising one fifth of the total number of villages. One set of
randomly selected village is chosen for crop inspection on priority basis immediately after the
sowing in each season are completed, but in advance of the period prescribed in the land
records manuals for such crop inspection. The village crop area statements are submitted to
higher authorities in stipulated date to estimate crop area statistics in advance for major crops.
These estimates are further used for crop forecasting purposes. The sampled villages under
TRS are selected from temporary area in such a way that the entire temporarily settled parts
of the country are covered over a period of five years.
The TRS provides for recording the area under irrigation as well as area under high yielding
varieties in the selected villages. Besides ensuring accuracy and timeliness of the
enumeration of the area under crops, statistical staff under the scheme is required to inspect
the fieldwork of crop cutting experiments and ensure timely dispatch of the returns. This
scheme has been taken up in a phased manner in different States beginning with Uttar
Pradesh and Maharashtra.
4. Establishment of an Agency for Reporting of Agricultural Statistics Scheme (EARAS)
In the states of Kerala, Orissa and West Bengal a scheme similar to TRS was introduced with
same objectives of obtaining area estimates based on 20% sample for use of both by Center
and States. Here also, it was envisaged that complete enumeration of fields for area figures
would be available for all villages over a period of five years as in case of TRS.
5. System of Data Collection for Area Estimation
In the states where land record are maintained (temporary settled) the village accountant is
in-charge of a village or a group of villages for carrying out field to field crop inspection in
each crop season for an agricultural year to record the crop area and land utilization statistics.
He is supposed to record the crop details related to area and land utilization in Khasra
register. After the completion of entries for each survey number of the village, an abstract of
area sown under different crops “Jinswar statement” is prepared and sent to next higher
official in the revenue hierarchy. At the end of each agricultural year a land utilization area
statistics are compiled and abstract is sent to related higher official. The crop wise and land
utilization wise area statistics obtained from different villages are aggregated at the revenue
circle, tehsil and district levels. The district wise area statistics are sent to State Agricultural
Statistics Authority (SASA), which is generally Director of Statistical Bureau or the Director
of Agriculture or the Director of Land Records. The state level aggregation is done by SASA
and forwarded to Directorate of Economics and Statistics (DES), Ministry of Agriculture &
Cooperation, Govt. of India, which is the nodal agency for releasing the state level and the all
India level estimates.
In order to improve the timeliness and quality of Agricultural Statistics, Ministry of
Agriculture and Cooperation, Govt. of India introduced TRS and EARAS. Area enumeration
under TRS has to be completed on priority basis in a random sample of 20% of the villages
during each crop season in a state. EARAS was introduced as a sequel to TRS in the non-land
record states namely Kerala, Orissa and West Bengal. This scheme provides for setting up
whole time agency to cover 20% of villages every year so that all the villages of a state are
covered in 5 years. In the sample villages under this scheme, the crop area is to be reported
on the basis of complete enumeration.
I.3
6. System of Data Collection for Yield Estimation
The sampling design generally adopted for the Crop Estimation Surveys is one of Stratified
Multi-Stage Random Sampling with tehsils/taluks/revenue inspector circles/blocks/anchals,
etc. as strata and revenue village within a stratum as first stage unit of sampling, survey
number/field within each selected village as sampling unit at the second stage and
experimental plot of a specified shape and size as the ultimate unit of sampling. In each
selected primary unit generally two survey numbers/fields growing the experimental crop are
selected for conducting crop-cutting experiments. However, in Dadra and Nagar Haveli three
fields are selected instead of two.
Generally, 80-120 experiments are conducted for a crop in a major district where a district is
considered as major for a given crop if the area under the crop in the district exceeds 80,000
hectares or lies between 40,000 and 80,000 hectares but exceeds the average area per district
in the State. Otherwise, district is considered a minor for a given crop. Experiments in minor
districts are so adjusted that the precision of the estimates is fairly high and the workload on
the field staff is manageable. On an average, about 44 or 46 experiments are planned in a
minor district. The number of experiments allotted to a district is distributed among the strata
within the district roughly in proportion to the area under the crop in the stratum. Generally,
the crop cutting is done in a plot of size 5m x 5m size for most of the crops in most of the
states. However, in UP the shape of the plot is of an equilateral triangle of size 10 meters and
in West Bengal a circular plot of radius 1.745 meters is taken for crop cut.
The average yield is obtained after harvesting, threshing, weighing and recording the weight
of the produce from the selected plots. In a sub-sample of experiments further processing of
the harvested produce is done to determine the percentage recovery of dried grains or the
marketable grain of the produce depending on the nature of the crop.
In the case of three non-land record states i.e. Kerala, Orissa and West Bengal both area and
yield are estimated on the basis of sample surveys. The crop cutting experiments are planned
in a sub-sample of the primary units selected for the purpose of area enumeration. The
general procedure of selecting sampling units remains same at different stages as in that of
other states. However, some special features of these states need to be mentioned specifically.
In Kerala block/city corporation or municipalities with an area of 10 sq. km. and above are
treated as separate stratum. Municipalities with an area of less than sq. km. are merged with
adjoining blocks and treated as a single stratum. These blocks are divided into a number of
Investigator Zones depending on the area of a block, nature of land, etc. City Corporation
area is divided into three Investigator Zones. Each municipality with an area more than 10 sq.
km. is treated as a single Investigator Zone. The number of crop cutting experiments
conducted in each Investigator Zone is six per season for paddy, three each for Coconut and
Banana and two each for Tapioca, Arecanut, Cashewnut, Pepper, Plantain and Jackfruit in an
agricultural year. In a municipal area having separate Investigator Zone, 10 crop-cutting
experiments are conducted in respect of paddy per season and 5 for Coconut per year. For
City Corporation areas, six experiments for paddy per season and five for coconut per year in
one Investigator Zone are conducted.
7. System for Crop Forecasting
The advance estimates of crop area and production are released with respect to principal food
and non-food crops (food grain, oilseeds, sugarcane, fibres etc.), which covers nearly 87% of
agricultural output. Four forecasts are issued, first in middle of September, the second in
January, the third towards the end of March and fourth by the end of May.
I.4
The advance estimates released in September are related to Kharif crops, which is mostly
based on reports submitted by states based on visual observation of the field officials. The
second forecast which covers both Kharif and Rabi and released in January by taking into
account additional information obtained from various sources including agricultural inputs,
incidence of pests and diseases, weekly reports from state government regarding area
coverage, conditions of standing crops etc. Presently estimates obtained through Remote
Sensing are also considered at this stage. The third forecast, which is made in March, the
estimates of Kharif and Rabi seasons are revised based on information received from sources
such as Market Intelligence Units, Meteorological Department and the Crop Weather Watch
Group (CWWG). The forecast made by the end of May is based on actual figures supplied by
State Agricultural Statistics Authorities (SASAs) using yield estimates obtained through
GCES. In addition to these four forecasts, the DES, MOA provides final estimates in
December. The fully revised estimates are obtained in the next crop year in the following
December in which all delayed information are incorporated and all India crop statistics are
released.
The National Crop Forecasting Centre (NCFC) was setup by Ministry of Agriculture with the
objective of examining existing mechanism of making forecasts and developing more
objective technique. However, the NCFC need to strengthen the crop forecasting system of
the country by incorporating more objective techniques and models based on sound statistical
techniques.
8. Co-ordination of Data Collection
The Field Operation Division of NSSO has the overall responsibility of assisting the States in
developing suitable techniques for obtaining reliable and timely estimates, providing
technical guidance and ensuring adoption of uniform concepts, definitions and procedures in
the Crop Estimation Survey (CES) in the States. It reviews the design, plan, details of
implementation and the results of the surveys and, associates itself in the conduct of training
camps organized for the States field staff and participates in the primary field work of
exercising technical supervision.
9. Supervision of Data Collection
Supervision of fieldwork is an essential part of any large-scale sample survey for ensuring
quality of data collected. A three fold approach is adopted in the States for supervision of
crop cutting experiments planned under Crop Estimation Surveys. This includes :
a. Supervision by the statistical staff of State Agricultural Statistics Authorities (SASAs),
b. Supervision by the Departmental staff i.e. by the supervisory officers of the
Departments whose workers are responsible for the conduct of crop cutting experiments
in the field and
c. Supervision by the Technical personnel of the FOD of National Sample Survey Office.
In the States of Goa, Orissa, West Bengal and the UT of Pondicherry where the field work
was conducted only by the staff of Statistics Department, the supervision was done by the
Statistical staff only whereas in the case of Bihar, Himachal Pradesh, Union Territories of
Dadra & Nagar Haveli and Daman & Diu, though there are other primary field agencies, the
supervision was done by the State statistical staff only. Though supervision of the conduct of
Crop Cutting Experiments in various states was in vogue since inception of Agricultural
Statistical Wing from the year 1973-74 (Rabi) onwards, NSSO personnel are participating in
the supervision by conducting sample check on crop cutting experiments in the post-harvest
stages in a pre-assigned sample under the Scheme for Improvement of Crop Statistics (ICS)
I.5
in 20 states and 2 Union Territories. Under this Scheme, State statistical staff also undertakes
similar sample checks on a matching basis.
10. Applications of Remote Sensing and GIS Technology
In India, Indian Council of Agricultural Research (ICAR) and Indian space Research
Organization (ISRO) jointly conducted the first multi-spectral air born study for identification
of root-wilt disease in coconut in 1969.
The country level studies related to applications of remote sensing technologies were initiated
after launch of IRS-IA satellite. Crop Acreage and Production Estimation (CAPE) was one of
the important projects in this direction for estimation of crop area under wheat, rice, cotton,
ground nut, sorghum & mustered. Apart from these national level projects, numbers of small
studies have been carried out to develop methodologies for application of satellite data in
various fields of agricultural and rural development by Department of Space. Some of these
studies are by Dadhwal et al. (1985,1991), etc.. Several methodological studies related to
estimation of crop area and production have been carried out at Indian Agricultural Statistics
Research Institute (IASRI), New Delhi. Singh et al. (1992) used satellite data for
stratification of crop area for the general crop estimation surveys and obtained more precise
estimator of crop yield. Singh et al. (1999) also developed small area estimator of crop yield.
Singh et al. (2002) used satellite data and the farmers eye estimate for developing a reliable
crop yield model. Application of remote sensing and GIS technology for estimation of land
use statistics using spatial models has been explored by Rai et al. (2004). Now, a project
entitled “Forecasting Agricultural Output Using Space, Agro-metrology and Land-based
Observations (FASAL) is undertaken under National Crop Forecasting Center (NCFC) of
Ministry of Agriculture, to meet the requirements of timely nation wide and multi- crop
reliable fore cast. A project has also been taken up jointly by IASRI New Delhi, Space
Application Center (SAC) Ahamdabad and North-Eastern Space Application Centre
(NESAC) Shillong with the support of Directorate of Economic and Statistic of Meghalya
State to explore the possibility of estimation of area and production of field crops by
integration of remote sensing technology, GIS and field survey. The result of all these studies
are very encouraging and indicates that in future remote sensing and GIS has a great potential
tool to improve the quality of area and production statistics of the country.
11. Improvement of Crop Statistics (ICS)
In addition to the crop area estimates developed by the state government the National Sample
Survey (NSS) use to develop area estimates based on sample surveys during its regular
rounds of surveys. Estimates were obtained for the whole country and also for certain
population zones. There used to be significant differences between two series of data on crop
area statistics. In order to probe into these high differences a technical committee on crop
statistics was set up in 1963. The committee favoured inter alia the estimates based on
complete enumeration. As a consequence the NSS discontinued its land utilization surveys
and also crop cutting experiments in 1970-71 under household surveys. Thereafter, the NSSO
introduced the ICS scheme in 1973-74 with an objective of improving the quality of statistics
through joint efforts of centre and state authorities. Currently the scheme is in operation in
20 states and two Union Territories of Delhi and Pondicherry. In this scheme an independent
agency (NSSO) carries out the supervision and physical verification of girdawari in a sub-
sample of four clusters of five survey numbers in each of the TRS sample villages. An
assessment is made for extent of discrepancies between the entries of supervisor and
girdawari completed by village accountant for each of the selected survey numbers in the
sample. The supervisors for checking possible errors of aggregations also scrutinize the crop
abstract of the village, which is prepared by patwaries. The permanently settled states are also
I.6
covered under this scheme where a sub-sample of EARAS sample villages (survey number)
is scrutinized following the same methodology as adopted for temporarily settled area.
Generally, a total of 10,000 sample villages are covered by the ICS out of which 8,500 are in
the temporarily settled states and 1,500 in the permanently settled states.
National Sample Survey Office (NSSO) is mainly responsible for planning and operations of
ICS by employing full time field staff for supervision. The responsibility of field supervision
is shared by designated state agencies which are responsible for carrying out the field
supervision in approximately half of the sampled villages.
11.1 Major Issues Emerging from the ICS Scheme
1. The crop statements submitted by patwari are many times based on incomplete
girdawaries.
2. The village crop statements are not submitted in time and there are large percentages
of non-response.
3. The entries in the girdawaries are not correct at least for one third of survey numbers.
4. Recording area under mixed crop is a major source of errors as it is not uniform across
the states.
5. Sometimes there is uncertainty of recording area under crop as area sown or area
harvested. This leads to inaccurate estimation of area, if area sown is recorded as area
under crop and there is no germination as expected.
6. Area sown more than once is also responsible for some confusion about statistics of
area under various crops.
7. Inclusion of field ridges, bunds in measurements also result in accuracy, which may be
higher in some of the cases.
8. Due to introduction of new technology/varieties number of short duration crops are
grown and also, there is shift in cropping pattern towards value added crops which are
not reflected properly in girdawari.
9. It has been observed that field staff approved by the State Government do not strictly
adhere to the prescribed procedures and thereby the survey estimates are subject to a
variety of non-sampling errors.
10. The errors are introduced mainly due to wrong selection of fields and duration of
selected experimental plots. The use of defective instruments such as proper weighing
machine introduces considerable amount of measurement errors.
11. The state departments of revenue and agriculture, which are responsible for carrying
out the survey, keeps these programmes on low priority and there is inadequate higher
level of supervision and control of field operations. The “High Level Coordination
Committee (HLCC) on Agricultural Statistics” in the states has little impact in
improving the quality of data.
12. In order to meet the requirements of getting estimates at block/village panchayat
levels especially for crop insurance purposes some of the State increased the number
of crop cutting experiments considerably. This imposes an enormous burden on the
field agency, increases considerably the non-sampling errors, which results in further
deterioration of quality of data collected through GCES. There is possibility of under
estimation of yield rates in case of crop insurance due to local pressure from insured
farmers where interest lies in depressing the crop yield.
I.7
13. It has been a matter of great debate in the past as production statistics obtained by
different sources/agencies are quite different. The problem is especially significant in
case of cash crops like cotton, oilseeds etc.
14. Inadequate training is provided to the field staff for conducting the crop cutting
experiments.
15. Another important factor, which has bearing on the quality of production data is, the
late time schedule fixed for certain crops in Kharif season in some states. In this case
crop-cutting experiments are to be conducted before completion of the season due to
early harvesting. Such situations have been arising in respect of Kharif crops like
maize, jowar, bajra, groundnut, cotton, soyabean etc. in States like Gujarat, Haryana,
Karnataka and M.P. Due to early harvesting of these crops, area under crop is
generally under reported and hence production too.
16. There is strong need to develop suitable forecasting models which integrate
information from different sources on parameters related to crop production such as
crop conditions, agro meteorology, water availability etc.
17. No multi-dimensional models exists in which the information generated from different
sources can be integrated.
18. The flows of information from different generating agencies are not time bound and
appropriate.
19. The DES, MOA is loosing confidence of users group due to frequent changes in
production figures specially most of the time differences in the forecasted estimates
are huge. These differences create lot of confusion and doubt among users
20. The present technique is mostly subjective and is not based on sound statistical
technique.
References
1. Basu, D. (1969). Role of sufficiency and likelihood principles in sample survey
theory. Sankhya, 31, 441-454.
2. Basu, D. (1971). An essay on the logical foundations of survey sampling, Apart I. In
Foundations of Statistical Inference, Holt, Rinehast and Winston, Toronto, 203-242.
3. Bowley, A.L. (1906). Address to the Economic Science and Statistics Section of the
British Association for the Advancement of Science. J. Roy. Statist. Soc., 69, 548-557
4. Bowley, A.L. (1926). Measurement of the precision attained in Sampling. Bull. Int.
Statist. Inst., 22 Livre I.
5. Brewer, K.R.W. (1963). A model of systematic sampling with unequal probabilities.
Austral. J. Statist., 5, 3-5.
6. Cassel, C.M., Sarndal, C.E., and Wretman, J.H. (1977). Foundations of Inference in
Survey Sampling. New York: Wiley.
7. Cochran, W.G. (1942). Sampling theory when the sampling units are of unequal
sizes.’ J. Amer. Statist. Assoc., 37, 199-212.
8. Cochran, W.G. (1953). Sampling Techniques, 1st ed., 2nd ed. (1963). 3rd ed., (1977).
New York: Wiley.
9. Cochran, W.G. and Watson, D.J. (1936). An experiment on observer’s bias in the
selection of shoot-heights. Empire J. Exp. Agriculture, 4, 69-76.
I.8
10. Datenius, T. (1962). Recent advances in sample survey theory and methods. Ann.
Math. Statist. 33, 325-349.
11. Deming, W.E., (1950). Some Theory of Sampling. New York. Wiley
12. Godambe, V.P. (1955). A unified theory of sampling from finite populations. J. Roy.
Statist. Soc. B, 17, 269-278.
13. Hansen, M.H.., Hurwitz, W.N. and Madow, W.G. (1953). Sample survey methods
and theory. John Wiley and Sons, New York, Vols. I and II.
14. Horvitz, D.G. and Thompson, D.J. (1952). A generalisation of sampling without
replacement from a finite universe. J. Amer. Statist. Assoc., 47, 663-685.
15. Jessen, R.J. (1942). Statistical investigation of a sample survey for obtaining farm
facts. Iowa Agricultural Experimental Station Research Bulletin No. 304.
16. Kiaer, A.N. (1895-6). Observations of experiences concernant des denombrements
representatifs. Bull. Int. Statist. Inst., 9, Liv. 2, 176-183.
17. McCarthy, P.J. (1966). Pseudo-replication: an approach to the analysis of data from
complex surveys. Washington: NCHS Series 2, No.14.
18. Narain, R.D. (1951). On sampling without replacement with varying probabilities, J.
Ind. Soc. Agril. Statist., 3, 169-174.
19. Neyman, J. (1934). On the two different aspects of the representative method: The
method of stratified sampling and the method of purposive selection. J. Roy. Statist.
Soc., 97, 555- 606.
20. Neyman, J. (1938). Contribution to the theory of sampling human populations. J.
Amer. Statist. Assoc., 33, 101-116.
21. O’Muircheartaigh, C. and Wong, T. (1981). The impact of sampling theory on survey
sampling practice. Bull. Int. Statist. Inst. ILIX Book 1, 465-493.
22. Patterson, H.D. (1950). Sampling on successive occasions with partial replacement of
units. J. Roy. Statist. Soc., B, 312, 241-255.
23. Rao, C.R., (1985): Evaluation of data collection censuses. Sample Surveys and Design
of Experiments.
24. Rao, J.N.K. (1963). On two systems of unequal probability sampling without
replacement. Ann. Inst. Statist. Maths. 15, 67-72.
25. Rao, J.N.K. and Scott, A.J. (1981). The analysis of categorical data from complex
sample surveys: Chi-squared tests for goodness of fit and independence in two-way
tables. J. Amer. Statist. Assoc., 76, 221-230.
26. Royall, R.M. (1968). An old approach to finite population sampling theory. J. Amer.
Statist. Assoc., 63, 1269-1279.
27. Royall, R.M. (1970). On finite population sampling theory under certain linear
regression models. Biometrika, 57, 377-387.
28. Verma, V., Scott, C. and O’Muircheartaigh, C. (1980). Sample designs and sampling
errors for the World Fertility Survey. J. Roy. Statist. Soc., A143, 431-473.
29. Sukhatme, P.V. (1959). Major developments in the theory and application of
sampling during the last twenty five years. Estadistica, 17, 62-679.
I.9
HISTORICAL PERSPECTIVE OF SURVEY SAMPLING
A.K. Srivastava
Former Joint Director, I.A.S.R.I., New Delhi -110012
1. Introduction
The purpose of this article is to provide an overview of developments in sampling theory and
its application particularly with respect to survey data analysis. An attempt is made to present
a historical perspective and the background behind the major developments in the field of
sample surveys. It is interesting to observe that many of the questions raised and the concepts
visualized on intuitive arguments in the early stages of development have reappeared with
fresh theoretical justification. Some of these questions relate to foundational aspects of
sample surveys. The issues raised in this process have got strong bearing on analytical
aspects of survey data analysis. The article is based on various review papers available in
literature. However, due to vastness of the areas, attempt is concentrated only to the
important developments and exhaustiveness of the review is not attempted.
The developments in survey sampling may be classified into following four periods:
i. Early developments (Before Neyman (1934))
ii. Between 1934 to early 50’s (to be specific 1955)
iii. Between 1955 to early 80’s
iv. More recent developments (After 1980)
2. Early Developments
The first account of a strong plea for the use of samples in data collection was made by Kiaer
(1895) at I.S.I. meeting in Berne. Kiaer presented a report on his experience with sample
surveys conducted in the Norwegian Bureau of Statistics and advocated further investigations
in the field. Kiaer’s plea was received with skepticism. It was felt that the method of sample
survey could not replace the method of enumeration. At that stage it was felt that there were
three survey methods which were possible:
i. Complete enumeration
ii. Monography, and
iii. Statistical Exploration (term, then used to describe Kiaer’s method)
In the subsequent meetings at St. Petersburg (1899) and at Budapest (1901), Kiaer defended
his method emphasizing representativeness of the sample. Kiaer’s efforts bore fruits at the
Berlin Session of the ISI in 1903. In this Session, the ISI adopted a resolution which
recommended the use of the representative method, subject to the provision that in the
publication of the results the conditions under which the selection of the observations was
made were completely specified. The sample survey had become an acceptable method of
data collection. There are four important principles involved in Kiaer’s approach:
a) Representativeness
b) Lack of subjectivity
c) Reliability of the results should be assessed
d) Complete specification of the method of selection be included with the results of any
sample survey
It may well be realized that how important these considerations have been in shaping the
future of sampling theory and its application. Kiaer’s method of selection had been in effort,
a proportional stratified multistage sampling (of course without random selection). Bowley
(1906) was the first to supply a theory of inference for survey samples and using Edgeworth’s
Bayesian version of the Central Limit Theorem. He was able to assess the accuracy of
estimators made from large samples drawn by simple random sampling from large finite
populations. His theoretical analysis showed that very often quite small samples are good
enough and census was not always necessary. Bowley’s method was, however, limited to
simple random sampling only. At this stage, random sampling was considered a feasible
proposition and was treated at par with purposive method of selection.
The two methods represent logical developments of the methodology presented by Kiaer. The
method of random selection carried to its logical conclusion that the selection of units should
be made in an objective manner. The method of purposive selection was a logical extension
of the principle that the sample should be a miniature of the whole population. What was
lacking was the means to bring these two principles together.
3. Neyman (1934) and Subsequent Developments
Neyman’s (1934) paper ‘On the two different aspects of the representative method: the
method of stratified sampling and the method of purposive selection’ was a watershed in the
development of sampling theory and its practice. His paper was a powerful and convincing
thesis which established random sampling not only as a viable alternative but also as a much
superior tool than purposive method of selection. He developed a theory of inference based
on confidence intervals which is suitable for use with populations of kind encountered in
survey work. He suggested that it was possible; using the idea of confidence intervals, to
define what he termed as a representative method of sampling and a consistent method of
estimation.
The use of confidence intervals derived from the probabilities due to the sampling design
provided a new framework of inference for finite population sampling. No longer was there
any need to make prior assumptions about the population as was necessary for the application
of Bayes’ argument. The validity of confidence statements did not depend on any
assumptions about the population. This new framework of inference was basically non-
parametric and, as such particularly attractive. This feature was, in fact, so appealing that for
the next thirty years or so, inference for samples drawn from infinite populations were firmly
tied to the distribution generated by the randomization of the sampling design. This
distribution is sometimes called the p-distribution.
Neyman, considering the choice of estimators which could be most suitable for the
construction of confidence intervals, suggested that two requirements could be formulated as
follows:
I.11
i. They must follow a frequency distribution which is already tabulated or may be easily
calculated
ii. The resulting confidence intervals should be as narrow as possible
The first of these requirements was suggested because of practical considerations. The second
introduced the concept of efficiency. Neyman’s final conclusions could be summarized as
follows:
i. It is possible to select random (probability) samples of groups of units (clusters)
which are measurable (i.e. the variance can be estimated from the sample itself)
ii. The number of sampling units in the sample should be large;
iii. Subjective or objective information can be used in sample design without departing
from probability sampling; to make the validity of the estimates dependent on prior
guesses about the population is dangerous and generally unsuccessful.
iv. The only method which can be advised for general use is stratified random sampling
v. There are instances where purposive selection may be used with success, but such
instances are exceptional and should not be used to determine the general strategy of
sample design.
The next 15 years (1935 to 1950) witnessed a very rapid all round growth in survey sampling
based on random sampling approach. Various probability sampling methods were developed
and refined during this period. A theory for systematic sampling, which was already used in
early stages, was developed (Madow and Madow, 1944). Probability proportional to size
sampling method was introduced (Hansen and Hurvitz (1943). Ratio and regression method
of estimation were also introduced during the 1930 with a comprehensive account of the
theory being provided by Cochran (1942). The concept of double sampling was introduced by
Neyman (1938). Sampling over successive occasion was introduced by Jessen (1942) and
further developed by Patterson (1950). An interesting feature of the developments during this
period was that methods developed were being simultaneously tested through actual surveys.
In fact, the need for various methods was coming from practical consideration only. In India
at Indian Statistical Institute, Calcutta, various methods like Interpenetrating sampling and
cost functions were developed. The jute survey conducted by Mahalanobis (1938) is an
elegant example for conducting pilot sample surveys. The methodology of crop estimation
surveys through crop cutting techniques was developed at Indian Council of Agricultural
Research. By 1950 quite substantial developments had taken place which were consolidated
in the form of various text books (Yates (1949), Deming (1950), Cochran (1953), Hansen,
Hurwitz and Madow (1953) and Sukhatme (1954).
4. Between 1955 to 1980
The concept of varying probability sampling which was considered for with replacement case
by Hansen, Hurwitz and Madow in 1943 was further developed for without replacement by
Narain (1951) and Horvitz and Thompson (1952). These papers led to a series of studies on
various methods of varying probability sampling without replacement. But a more important
impact was on future direction of research in the sampling theory. Horvitz and Thompson
considered three classes of linear estimators termed as T1, T2 and T3. These classes have now
been extended up to T8 classes. Godambe (1955), while investigating unified theory of
I.12
sampling, proved non-existence of uniformly minimum variance estimators. A more elegant
proof of this theorem was given by Basu (1971). In the choice of good estimators in a
particular class, various concepts like admissibility, hyperadmissibility etc. were considered.
Non-existence of the best estimator led to choice of best estimator in restricted classes. In the
absence of UMV estimators, concepts like necessary best estimators were considered.
Another direction in which the research in sampling theory progressed was attempts to bring
estimation problem in sampling theory closer to the estimation problem in usual statistical
inference. Concepts like sufficiency and likelihood function were considered. Sufficiency
concept lead to improvement of estimators through Rao-Blackwellisation of ordered
estimators and for estimators based on distinct units. One of the important finding which was
emerging from these considerations was that sampling design should not have a role to play
in the estimation of population parameters. In fact the likelihood function in sampling
problems is flat and, therefore, is non-informative. Simultaneously model based estimation
(Brewer (1963), Royall (1968, 70, 71, 72) were being developed. This approach was known
as predictive model of estimation. In this method also the estimation does not depend upon
the sampling design. In fact, it depends upon the model. A detailed discussion on these
developments is available in the book by Cassel, Sarndal and Wretman (1977).
5. Analysis of survey data (1980 onwards)
During the last three decades besides the research regarding foundational and on inferential
aspects a change in emphasis has been towards the analysis of survey data such as
contingency tables of estimated counts, logistic regression and multivariate analysis. Further,
methods that take proper account of the complexity of survey data have been proposed. All
these became possible mainly due to enhanced computer capabilities. Most of the literature
on sampling theory deals with estimation of population parameters such as means, totals and
ratios along with their standard errors. In recent years, considerable attention has also been
given to complex descriptive parameters such as domain (subpopulation) totals and mean,
quantiles, regression and correlation coefficient.
Estimation of domain parameters i.e., the estimation of small area statistics has gained
importance in view of growing needs of microlevel planning. The main problem in the
estimation of domain parameters is that sample sizes in subpopulations are too small to
provide reliable estimates with the help of direct estimators. Hence, it becomes essential to
borrow information from related or similar areas through explicit and implicit models to
produce indirect estimator which increases the effective sample sizes. The advances in
computing facilities have also provided convenient tools for many theoretical developments
in this area.
Assumption of classical sampling theory methods i.e., units of the sample are independently
and identically distributed (i.i.d), no longer hold in case of complex survey data as these data
are collected through complex sampling designs. For these purposes even sampling designs
like stratified and cluster sampling may be treated as complex ones.
Regression coefficients are estimated with the help of multiple regression technique which
assumes the independence of observations. Consequently, use of standard ordinary least
squares (OLS) technique to survey data for estimating regression coefficient provides
misleading result due to dependent sample units. Efforts have been made to incorporate the
effect of sampling design with the help of various approaches of inferences.
I.13
With the inexpensible and easy access of computers, researchers are interested not only in the
descriptive surveys but also in the analytical surveys i.e., investigating relationships among
variables in the survey. The analysis of categorical variable from survey data comes under
analytical survey. Here also due to violation of i.i.d assumption, standard statistics such as
chi-square tests need adjustment to ensure valid conclusions.
Lastly, the area of variance estimation has also gained importance in view of complex survey
design as well as vast potential of modern computation. The traditional approach need to
derive a theoretical formula of variance for each estimator which is quite cumbersome in case
of complex survey designs. In case of non-linear estimators it may not even be possible
sometimes to obtain such a formulae. However, if we take recourse to what are called
Resampling Techniques, we need only to obtain and use single and simple expressions for
estimator of variance for all the estimators under the given sampling design. Of course, these
techniques are highly computer intensive.
To conclude it may be remarked that analysis of survey data in the large scale surveys
conducted in the country has been mainly on traditional lines. Utilisation of data through
various aspects of recent approaches of complex data analysis are yet to be made. In fact,
such attempts will go a long way towards proper utilization of vast amount of data being
collected in the country in a regular way.
References
1. Basu, D. (1971). An essay on the logical foundations of survey sampling, Apart I. In
Foundations of Statistical Inference, Holt, Rinehast and Winston, Toronto, 203-242.
2. Bowley, A.L. (1906). Address to the Economic Science and Statistics Section of the
British Association for the Advancement of Science. J. Roy. Statist. Soc., 69, 548-557.
3. Brewer, K.R.W. (1963). A model of systematic sampling with unequal probabilities.
Austral. J. Statist., 5, 3-5.
4. Cassel, C.M., Sarndal, C.E., and Wretman, J.H. (1977). Foundations of Inference in
Survey Sampling. New York: Wiley.
5. Cochran, W.G. (1942). Sampling theory when the sampling units are of unequal
sizes.’ J. Amer. Statist. Assoc., 37, 199-212.
6. Godambe, V.P. (1955). A unified theory of sampling from finite populations. J. Roy.
Statist. Soc. B, 17, 269-278.
7. Hansen, M.H., Hurwitz, W.N. and Madow, W.G. (1953). Sample survey methods and
theory. John Wiley and Sons, New York, Vols. I and II.
8. Horvitz, D.G. and Thompson, D.J. (1952). A generalisation of sampling without
replacement from a finite universe. J. Amer. Statist. Assoc., 47, 663-685.
9. Jessen, R.J. (1942). Statistical investigation of a sample survey for obtaining farm
facts. Iowa Agricultural Experimental Station Research Bulletin No. 304.
10. Kiaer, A.N. (1895-6). Observations of experiences concernant des denombrements
representatifs. Bull. Int. Stat. Inst., 9, Liv. 2, 176-183.
I.14
11. Narain, R.D. (1951). On Sampling without replacement with varying probabilities, J.
Ind. Soc. Agril. Statist., 3, 169-174.
12. Neyman, J. (1934). On the two different aspects of the representative method: The
method of stratified sampling and the method of purposive selection. J. Roy. Statist.
Soc., 97, 555- 606.
13. Neyman, J. (1938). Contribution to the theory of sampling human populations. J.
Amer. Statist. Assoc., 33, 101-116.
14. Royall, R.M. (1968). An old approach to finite population sampling theory. J. Amer.
Statist. Assoc., 63, 1269-1279.
15. Royall, R.M. (1970). On finite population sampling theory under certain linear
regression models. Biometrika, 57, 377-387.
16. Sukhatme, P.V. (1959). Major developments in the theory and application of
sampling during the last twenty five years. Estadistica, 17, 62-679.
I.15
OVERVIEW OF SAMPLING SCHEMES
Yogita Gharde
I.A.S.R.I., New Delhi -110012
1. Introduction
The need to gather information arises in almost every conceivable sphere of human activity.
Many of the questions that are subject to common conservation and controversy require
numerical data for their resolution. The data collected and analyzed in an objective manner
and presented suitably serve as basis for taking policy decisions in different fields of daily
life.
The important users of statistical data, among others, include government, industry, business,
research institutions, public organizations and international agencies and organizations. To
discharge its various responsibilities, the government needs variety of information regarding
different sectors of economy, trade, industrial production, health and mortality, population,
livestock, agriculture, forestry, environment and available resources. The inferences drawn
from the data help in determining future needs of the nation and also in tackling social and
economic problems of people. For instance, the information on cost of living for different
categories of people, living in various parts of the country is of importance in shaping its
policies in respect of wages and price levels. Data on agricultural production are of immense
use to the state for planning to feed the nation. In case of industry and business, the
information is to be collected on labour, cost and quality of production, stock and demand
and supply positions for proper planning of production levels and sales campaigns.
1.1 Complete Enumeration
One way of obtaining the required information at regional and country level is to collect the
data for each and every unit (person, household, field, factory, shop etc. as the case may be)
belonging to the population which is the aggregate of all units of a given type under
consideration and this procedure of obtaining information is termed as complete enumeration.
The effort, money and time required for the carrying out complete enumeration to obtain the
different types of data will, generally, be extremely large. However, if the information is
required for each and every unit in the domain of study, a complete enumeration is clearly
necessary. Examples of such situations are preparation of “voter list” for election purposes
and recruitment of personnel in an establishment etc. But there are many situations, where
only summary figures are required for the domain of study as a whole or for group of units.
1.2 Need for Sampling
An effective alternative to a complete enumeration can be sample survey where only some of
the units selected in a suitable manner from the population are surveyed and an inference is
drawn about the population on the basis of observations made on the selected units. It can be
easily seen that compared to sample survey, a complete enumeration is time-consuming,
expensive, has less scope in the sense of restricted subject coverage and is subject to greater
coverage, observational and tabulation errors. In certain investigations, it may be essential to
use specialized equipment or highly trained field staff for data collection making it almost
impossible to carry out such investigations. It is of interest to note that if a sample survey is
carried out according to certain specified statistical principles, it is possible not only to
estimate the value of the characteristic for the population as a whole on the basis of the
sample data, but also to get a valid estimate of the sampling error of the estimate. There are
various steps involved in the planning and execution of the sample survey. One of the
principal steps in a sample survey relate to methods of data collection.
1.3 Methods of Data Collection
The different methods of data collection are:
1) Physical observation or measurement
2) Personal interview
3) Mail enquiry
4) Telephonic enquiry
5) Web-based enquiry
6) Method of Registration
7) Transcription from records
The first six methods relate to collection of primary data from the units/ respondents directly,
while the last one relates to the extraction of secondary data, collected earlier generally by one or
more of the first six methods. These methods have their respective merits and therefore sufficient
thought should be given in selection of an appropriate method(s) of data collection in any survey.
The choice of the method of data collection should be arrived at after careful consideration of
accuracy, practicability and cost from among the alternative methods.
Physical observations or measurement
Data collection by physical observation or measurement consists in physically examining the
units/respondents and recording data as a result of personal judgment or using a measuring
instrument by the investigator. For instance, in a crop cutting experiment for estimating the yield
of a crop, the plot is demarcated, the crop in the selected plot is harvested and the produce is
weighted to estimate the produce per unit area. Data obtained by this method are likely to be
more accurate but may often prove expensive.
Personal Interview
The method of personal interview consists in contacting the respondents and collecting statistical
data by questioning. In this case, the investigator can clearly explain to the respondents the
objectives of the survey and the exact nature of the data requirements and persuade them to give
the required information, thus reducing the possibility of non-response arising from non-
cooperation, indifference etc. Further, this method is most suitable for collecting data on
conceptually difficult items from respondents. However, this method depends heavily on the
availability of well trained interviewer.
Mail Enquiry
In a mail enquiry, data are collected by obtaining questionnaires filled in by the respondents, the
questionnaires being sent and collected back through an agency such as the postal department.
This method is likely to cost much less as compared to above methods. However, the response
may not always be satisfactory depending upon the cooperation of the respondents, the type of
questionnaire and the design of the questionnaire. In developing countries where a large
proportion of the population is illiterate, the method of mailed questionnaire may not even be
feasible.
I.17
Telephonic enquiry
In telephonic enquiry, data are collected by questioning the respondents. This method provides
an opportunity of two-way communication and thus can reduce the possibility of item non-
response. However, this method can be used only for those surveys in which all units of target
populations have telephone otherwise it will cause bias in the results.
Web-based enquiry
The increasing popularity and wide availability of World Wide Web technologies provide a new
mode of data collection. In web-based enquiry, data are collected by obtaining questionnaires
filled in by the respondents, the questionnaires being posted on the net. One important advantage
of using computer technology in data collection is to minimize the loss of data owing to
incomplete or incorrectly completed data sets by using Client side validation. In an era of
information superhighway, this method is one of the fastest means of data collection. However,
in developing countries where a large proportion of the population does not have access to
Internet, the method of web-based enquiry may not serve the purpose for most of the surveys.
Various Internet sites are using this method for opinion poll on certain issues.
Method of Registration
In the registration method, the respondents are required to register the required information at
specified place. The vital statistics registration system followed in many countries provides an
illustration of the registration method. The main difficulty with this method, as in the case of the
mail enquiry, is the possibility of non-response due to indifference, reluctance, etc. on the part of
informants to visit the place of registration and supply the required data.
Transcription from records
The method of transcription from records is used when the data needed for a specific purpose are
already available in registers maintained in one or more places, making it no more necessary to
collect them directly from the original units at much cost and effort. The method consists in
compiling the required information from the registers for the concerned units. This method is
extensively used since a good deal of government and business statistics are collected as by-
product of routine administrative operations.
1.4 Various Concepts and Definitions
Element: An element is a unit about which we require information. For example, a field growing
a particular crop is an element for collecting information on the yield of a crop.
Population: It is the totality of elements under consideration on which inference is required.
Thus, all fields growing a particular crop in a region constitute a population. Populations may be
finite or infinite. A finite population contains countable number of elements. For example, fields
growing a particular crop in a region are an example of finite population. On the other hand
plants of a particular crop belonging to a large country like India is an example of infinite
population. Generally, sample surveys deal with finite populations.
Sampling units: A group of elements constitute a sampling unit. Elements belonging to different
sampling units are non-overlapping. A sampling unit may have one, more than one or sometimes
even no elements. Sampling units are convenient as well as relatively inexpensive to observe. For
example, it is convenient to select households for collecting data on milk produced by animals
rather than contacting the elements directly. Sampling units should be non-overlapping and
physically identifiable.
I.18
Sampling frame: An exhaustive list of all the sampling units constitutes a sampling frame. An
example of a sampling frame may be cultivator fields growing a particular crop or households
containing animals in a region.
Sample: A part of the population selected from a sampling frame for the purpose of making
inference about the population is called as a sample. For example, a subset of the cultivator fields
may be selected to estimate the yield of a crop in a region.
Sometimes complete enumeration or census is impracticable, impossible, to collect data on each
and every unit of a large population. The money, manpower and time required in observing all
the population units may be considerably high. Moreover, the chance of committing mistakes
increases when the volume of work is increased. Thus, a sample survey (only some of the units
of the population are selected in a suitable manner) may be a better alternative for making
inferences about large populations.
Probability Sampling: Probability sampling is a procedure where the units in the sample are
selected by a probability mechanism. Thus, each unit in the population is assigned a
predetermined probability of being selected in the sample. The procedure assigns every possible
sample a known probability of selection. It is possible to get the frequency distribution of
samples according to values of the estimator based on the selected sample. Further, it is possible
to determine what proportion of values of estimators will lie in a given interval around the true
value.
Non-Probability Sampling: Non-Probability Sampling is the sampling procedure in which the
units are selected according to the choice of sampler. Thus, the choice of selection of sampling
units entirely depends upon the judgement of the sampler. First he observes and inspects the
sampling units and selects a suitable sample which he considers the most likely to the population
or representative of the population.
This sampling procedure is only used for opinion surveys and not for general survey because it is
a biased method of sampling and it is not possible to obtain the precision of the estimate from the
sample values.
Sampling Error: The error which arises due to use of sample to estimate the population
parameters is called as sampling error. Whatever method of sampling is used, there will always
be a difference between population value and its corresponding estimate.
This error is unavoidable in every sampling scheme. A sample with the smallest sampling error
will always be considered as a good representative of the population. This error can be reduced
by increasing the size of the sample. Thus, when the sample survey becomes a census or
complete enumeration, the sampling error becomes zero.
Non-Sampling Error: Besides sampling error, the sample estimate may be subject to other error
which arises due to failure to measure some of the units in the selected sample, observational
errors or errors introduced in editing, coding and tabulating the results.
Generally, census results may suffer from non-sampling error although these may be free from
sampling error. The non sampling error is likely to increase with increase in sample size, while
sampling error decreases with increase in sample size.
Population parameters: Suppose a finite population consists of the N units U1, U2,..., U N and
let Yi be the value of variable y, the characteristic under study, for the ith unit U i , (i=1,2,...,N).
For instance, the unit may be a farm and the characteristic under study may be the area under a
particular crop. Any function of the values of all the population units (or of all the observations
constituting a population) is known as a population parameter or simply a parameter. Some of the
I.19
N
important parameters usually required to be estimated in surveys are population total Y Yi
i 1
N
and population mean Y Yi / N .
i 1
I.20
The estimator can be expressed as
n n
Ŷ w i yi N / n yi , where w i N / n.
i 1 i 1
The constant N / n is the sampling weight and is the inverse of the sampling fraction n / N.
Alternatively, an estimator for the population total can be written by first defining the inclusion
probability of a population element. Under SRS, the inclusion probability of a population
element i is i = n/N, same or constant for every population element. Based on the inclusion
probabilities, an estimator of the total can be expressed as a more general Horvitz-Thompson-
type estimator
n n
1 N n
ŶHT w i yi
i 1
i 1 i
yi
n
y
i 1
i .
In this case, the estimator Ŷ and ŶHT obviously coincide, because the inclusion probabilities i =
n/N are equal for each i. The Horvitz-Thompson-type estimator is often used for example, with
probability-proportional to size sampling where inclusion probabilities vary. The estimator has
the statistical property of unbiasedness in relation to the sampling design. Variance of the
estimator Ŷ of the population total is given by
N2 n N
ˆ )
VSRS ( Y 1-
n N
i 1
(Yi Y)2 (N 1)
N N
where Y Yi / N is the population mean and S2 (Yi Y)2 /(N 1) is the population
i 1 i 1
mean square.
An unbiased estimator of variance of the estimator Ŷ of the total, VSRS( Ŷ ) is given by
ˆ ) N 2 1- n (y y ) 2 / n(n 1)
n
ˆ (Y
VSRS i n
N i 1
n
N 2 1- s2 / n
N
n
where yn yi / n is the sample mean and s2 is an unbiased estimator of the population mean
i 1
square S2.
3. Stratified Sampling
The basic idea in stratified random sampling is to divide a heterogeneous population into sub-
populations, usually known as strata, each of which is internally homogeneous in which case a
precise estimate of any stratum mean can be obtained based on a small sample from that stratum
and by combining such estimates, a precise estimate for the whole population can be obtained.
Stratified sampling provides a better cross section of the population than the procedure of simple
random sampling. It may also simplify the organization of the field work. Geographical
proximity is sometimes taken as the basis of stratification. The assumption here is that
geographically contiguous areas are often more alike than areas that are far apart. Administrative
convenience may also dictate the basis on which the stratification is made. For example, the staff
already available in each range of a forest division may have to supervise the survey in the area
under their jurisdiction. Thus, compact geographical regions may form the strata. If the
I.21
characteristic under study is known to be correlated with a supplementary variable for which
actual data or at least good estimates are available for the units in the population, the
stratification may be done using the information on the supplementary variable. For instance, the
volume estimates obtained at a previous inventory of the forest area may be used for
stratification of the population.
In stratified sampling, the variance of the estimator consists of only the ‘within strata’ variation.
Thus the larger the number of strata into which a population is divided, the higher, in general, the
precision, since it is likely that, in this case, the units within a stratum will be more
homogeneous. For estimating the variance within stratum, there should be a minimum of 2 units
in each stratum. The larger the number of strata the higher will, in general, be the cost of
enumeration. So, depending on administrative convenience, cost of the survey and variability of
the characteristic under study in the area, a decision on the number of strata will have to be
arrived at.
3.1 Allocation and Selection of the Sample within Strata
Assume that the population is divided into k strata of N1, N2,…,Nk units respectively, and that a
sample of n units is to be drawn from the population. The problem of allocation concerns the
choice of the sample sizes in the respective strata, i.e., how many units should be taken from
each stratum such that the total sample is n.
Other things being equal, a larger sample may be taken from a stratum with a larger variance so
that the variance of the estimates of strata means gets reduced. The application of the above
principle requires advance estimates of the variation within each stratum. These may be available
from a previous survey or may be based on pilot surveys of a restricted nature. Thus, if this
information is available, the sampling fraction in each stratum may be taken proportional to the
standard deviation of each stratum.
In case the cost per unit of conducting the survey in each stratum is known and is varying from
stratum to stratum an efficient method of allocation for minimum cost will be to take large
samples from the stratum where sampling is cheaper and variability is higher. To apply this
procedure one needs information on variability and cost of observation per unit in the different
strata.
Where information regarding the relative variances within strata and cost of operations are not
available, the allocation in the different strata may be made in proportion to the number of units
in them or the total area of each stratum. This method is usually known as ‘proportional
allocation’.
For the selection of units within strata, In general, any method which is based on a probability
selection of units can be adopted. But the selection should be independent in each stratum. If
independent random samples are taken from each stratum, the sampling procedure will be known
as ‘stratified random sampling’. Stratification, if properly done, will usually give lower variance
for the estimated population total or mean than a simple random sample of the same size.
4. Cluster Sampling
A sampling procedure presupposes division of the population into a finite number of distinct and
identifiable units called the sampling units. The smallest units into which the population can be
divided are called the elements of the population, and group of elements the clusters. A cluster
may be a class of students or cultivators’ fields in a village. When the sampling unit is a cluster,
the procedure of sampling is called cluster sampling.
For many types of population a list of elements is not available and the use of an element as the
sampling unit is, therefore, not feasible. The method of cluster or area sampling is available in
I.22
such cases. Thus, in a city a list of all the houses may be available, but that of persons is rarely
so. Again, list of farms are not available, but those of villages or enumeration districts prepared
for the census are. Cluster sampling is, therefore, widely practiced in sample surveys.
For a given number of sampling units cluster sampling is more convenient and less costly than
simple random sampling due to the saving time in journeys, identification and contacts etc., but
cluster sampling is generally less efficient than simple random sampling due to the tendency of
the units in a cluster to be similar. In most practical situations, the loss in efficiency may be
balanced by the reduction in the cost and the efficiency per unit cost may be more in cluster
sampling as compares to simple random sampling.
Clearly the size of the cluster will influence efficiency of sampling. In general, the smaller the
cluster, the more accurate will usually be the estimate of the population characteristic for a given
number of elements in the sample. Thus, a sample of farms independently and randomly selected
is likely to be scattered over entire area, and thereby provides a better cross-section of the
population than an equivalent sample, e.g. a sample of the same number of farms, clustered
together in a few villages. On the other hand, it will cost more to survey a widely scattered
sample of farms than to survey an equivalent sample of clusters of farms, since the additional
cost of surveying a neighboring farm is small as compared to the cost of locating a second
independent farm and surveying it. The optimum segment or cluster is one which would estimate
the characteristic under study with smallest standard error for a given proportion of the
population sampled, or more generally, for a given cost.
5. Multistage Sampling
Cluster sampling is a sampling procedure in which clusters are considered as sampling units and
all the elements of the selected clusters are enumerated. One of the main considerations of
adopting cluster sampling is the reduction of travel cost because of the nearness of elements in
the clusters. However, this method restricts the spread of the sample over population which
results generally in increasing the variance of the estimator. In order to increase the efficiency of
the estimator with the given cost it is natural to think of further sampling the clusters and
selecting more number of clusters so as to increase the spread of the sample over population.
This type of sampling which consists of first selecting clusters and then selecting a specified
number of elements from each selected cluster is known as sub-sampling or two stage sampling,
since the units are selected in two stages. In such sampling designs, clusters are generally termed
as first stage units (fsu’s) or primary stage units (psu’s) and the elements within clusters or
ultimate observational units are termed as second stage units (ssu’s) or ultimate stage units
(usu’s). It may be noted that this procedure can be easily generalized to give rise to multistage
sampling, where the sampling units at each stage are clusters of units of the next stage and the
ultimate observational units are selected in stages, sampling at each stage being done from each
of the sampling units or clusters selected in the previous stage. This procedure, being a
compromise between uni-stage or direct sampling of units and cluster sampling, can be expected
to be (i) more efficient than uni-stage sampling and less efficient than cluster sampling from
considerations of operational convenience and cost, and (ii) less efficient than uni-stage sampling
and more efficient than cluster sampling from the view point of sampling variability, when the
sample size in terms of number of ultimate units is fixed.
It may be mentioned that multistage sampling may be the only feasible procedure in a number of
practical situations, where a satisfactory sampling frame of ultimate observational units is not
readily available and the cost of obtaining such a frame is prohibitive or where the cost of
locating and physically identifying the usu’s is considerable. For instance, for conducting a
socio-economic survey in a region, where generally household is taken as the usu, a complete
and up-to-date list of all the households in the region may not be available, whereas a list of
I.23
villages and urban blocks which are group of households may be readily available. In such a
case, a sample of villages or urban blocks may be selected first and then a sample of households
may be drawn from each selected village and urban block after making a complete list of
households. It may happen that even a list of villages is not available, but only a list of all tehsils
(group of villages) is available. In this case a sample of households may be selected in three
stages by selecting first a sample of tehsils, then a sample of villages from each selected tehsil
after making a list of all the villages in the tehsil and finally a sample of households from each
selected village after listing all the households in it. Since the selection is done in three stages,
this procedure is termed as three stage sampling. Here, tehsils are taken as first stage units
(fsu’s), villages as second stage units (ssu’s) and households as third or ultimate stage units
(tsu’s).
6. Systematic Sampling
In all other sampling methods, the successive units (whether elements or clusters) are selected
with the help of random numbers. But a method of sampling in which only the first unit is
selected with the help of random number while the rest of the units are selected according to a
pre-determined pattern, is known as systematic sampling. The systematic sampling has been
found very useful in forest surveys for estimating the volume of timber, in fisheries surveys for
estimating the total catch of fish, in milk yield surveys for estimating the lactation yield etc.
7. Successive Sampling
Many times surveys often gets repeated on many occasions (over years or seasons) for
estimating same characteristics at different points of time. The information collected on previous
occasion can be used to study the change or the total value over occasion for the character and
also in addition to study the average value for the most recent occasion. For example in milk
yield survey one may be interested in estimating the
1. Average milk yield for the current season,
2. The change in milk yield for two different seasons and
3. Total milk production for the year.
The successive method of sampling consists of selecting sample units on different occasions
such that some units are common with samples selected on previous occasions. If sampling on
successive occasions is done according to a specific rule, with partial replacement of sampling
units, it is known as successive sampling.
Generally, the main objective of successive surveys is to estimate the change with a view to
study the effects of the forces acting upon the population. For this, it is better to retain the same
sample from occasion to occasion. For populations where the basic objective is to study the total,
it is better to select a fresh sample for every occasion. If the objective is to estimate the average
value for the most recent occasion, the retention of a part of the sample over occasions provides
efficient estimates as compared to other alternatives.
8. Multiphase sampling
Multiphase sampling plays a vital role in forest surveys with its application extending over
continuous forest inventory to estimation of growing stock through remote sensing. The essential
idea in multiphase sampling is that of conducting separate sampling investigations in a sequence
of phases starting with a large number of sampling units in the first phase and taking only a
subset of the sampling units in each successive phase for measurement so as to estimate the
parameter of interest with added precision at relatively lower cost utilizing the relation between
I.24
characters measured at different phases. In order to keep things simple, further discussion in this
section is restricted to only two phase sampling.
A sampling technique which involves sampling in just two phases (occasions) is known as two
phase sampling. This technique is also referred to as double sampling. Double sampling is
particularly useful in situations in which the enumeration of the character under study (main
character) involves much cost or labour whereas an auxiliary character correlated with the main
character can be easily observed. Thus it can be convenient and economical to take a large
sample for the auxiliary variable in the first phase leading to precise estimates of the population
total or mean of the auxiliary variable. In the second phase, a small sample, usually a sub-
sample, is taken wherein both the main character and the auxiliary character may be observed
and using the first phase sampling as supplementary information and utilizing the ratio or
regression estimates, precise estimates for the main character can be obtained. It may be also
possible to increase the precision of the final estimates by including instead of one, a number of
correlated auxiliary variables. For example, in estimating the volume of a stand, we may use
diameter or girth of trees and height as auxiliary variables. In estimating the yield of tannin
materials from bark of trees certain physical measurements like the girth, height, number of
shoots, etc., can be taken as auxiliary variables.
9. Use of Auxiliary Information
In sampling theory if the auxiliary information, related to the character under study, is available
on all the population units, then it may be advantageous to make use of this additional
information in survey sampling. One way of using this additional information is in the sample
selection with unequal probabilities of selection of units. The knowledge of auxiliary information
may also be exploited at the estimation stage. The estimator can be developed in such a way that
it makes use of this additional information. Ratio estimator, difference estimator, regression
estimator, generalized difference estimators are the examples of such estimators. Obviously, it is
assumed that the auxiliary information is available on all the sampling units. In case the auxiliary
information is not available then it can be obtained easily without much burden on the cost.
Another way the auxiliary information can be used is at the stage of planning of survey. An
example of this is the stratification of the population units by making use of the auxiliary
information.
10. Sampling with Varying Probability
Under certain circumstances, selection of units with unequal probabilities provides more
efficient estimators than equal probability sampling, and this type of sampling is known as
unequal or varying probability sampling. In the most commonly used varying probability
sampling scheme, the units are selected with probability proportional to a given measure of size
(pps) where the size measure is the value of an auxiliary variable x related to the characteristic y
under study and this sampling scheme is termed as probability proportional to size sampling. For
instance, in estimating crop characteristics the geographical area or cultivated area for a previous
period, if available, may be considered as a measure of size, or in an industrial survey, the
number of workers may be taken as the size of an industrial establishment.
Since a large unit, that is, a unit with a large value for the study variable y, contributes more to
the population total than smaller units, it is natural to expect that a scheme of selection which
gives more chance of inclusion in a sample to larger units than to smaller units would provide
estimators more efficient than equal probability sampling. Such a scheme is provided by pps
sampling, size being the value of an auxiliary variable x directly related to y. In pps sampling,
the units may be selected with or without replacement.
I.25
References
1. Cochron, W.G. (1977). Sampling techniques. Wiley Eastern Ltd.
2. Des Raj, (1968). Sampling theory. Tata-Mcgraw-Hill Publishing Company Ltd.
3. Hansen, M.H. and Hurwitz, W.H. (1943). On the theory of sampling from finite
populations. Ann. Math. Statist., 14, 333-362.
4. Hansen, M.H., Hurwitz, W.H. and Madow, W.G. (1993). Sample survey methods and
theory. Vol. 1 and Vol. 2, John Wiley & Sons, Inc.
5. Murthy, M.N. (1977). Sampling theory and methods. Statistical Publishing Society.
6. Sukhatme, P.V., Sukhatme, B.V., Sukhatme, S. and Ashok, C. (1984). Sampling theory
of surveys with applications. Indian Society of Agricultural Statistics.
I.26
PLANNING AND ORGANIZATIONAL ASPECTS
OF SAMPLE SURVEYS
K.K. Tyagi
I.A.S.R.I., New Delhi -110012
1. Introduction
Sample surveys are widely used as a cost effective instrument of data collection and for
making valid inferences about population parameters. Most of the steps involved while
planning a sample survey are common to those for a complete enumeration. Three major
stages of a survey are planning, data collection and tabulation of data. Some of the important
aspects requiring attention at the planning stage are as follows:
1. formulation of data requirements - objectives of the survey
2. ad-hoc or repetitive survey
3. method of data collection
4. questionnaire versus schedules
5. survey, reference and reporting periods
6. problems of sampling frames
7. choice of sampling design
8. planning of pilot survey
9. field work
10. processing of data, and
11. preparation of report.
12. The different aspects listed above are inter-dependent.
(i) Formulation of Data Requirements
The users i.e. the persons or organizations requiring the statistical information are expected to
formulate the objectives of the survey. The user’s formulation of data requirements is not
likely to be adequately precise from the statistical point of view. It is for the survey
statistician to give a clear formulation of the objectives of the survey and to check up whether
his formulation faithfully reflects the requirements of the users. The survey statistician’s
formulation of data requirements should include the following:
i. a clear statement of the desired information in statistical terms
ii. specification of the domain of study
iii. the form in which the data should be tabulated
iv. the accuracy aimed at in the final results and
v. cost of survey
Besides, these aspects, one may accommodate some additional items of information, directly
or indirectly related to the objectives of the survey, which would provide checks on the
accuracy of data or assist in interpreting the results.
(ii) Survey: Ad-hoc or Repetitive
An ad-hoc survey is one which is conducted without any intention of or provision for
repeating it, whereas a repetitive survey is one, in which data are collected periodically for
the same, partially replaced or freshly selected sample units. If the aim is to study only the
current situation, the survey can be an ad-hoc one. But when changes or trends in some
characteristics over time are of interest, it is necessary to carry out the survey repetitively.
(iii) Methods of Collecting Primary Data
There are varieties of methods that may be used to collect information. The method to be
followed has to be decided keeping in view the cost involved and the precision aimed at. The
methods usually adopted for collecting primary data are:
i. Direct Personal Interview,
ii. Questionnaires sent through Mail,
iii. Interview by Enumerators, and
iv. Telephone Interview.
(a) Director Personal Interview
The method of personal interview is widely used in social and economic surveys. In these
surveys, the investigator personally contacts the respondents and can obtain the required data
fairly accurately. The interviewer asks the questions pertaining to the objective(s) of survey
and the information, so obtained, is recorded on a schedule prepared for the purpose. This
method is mostly suitable for collecting data on conceptually difficult items from
respondents. Under this method, the response rate is usually good and the information is more
reliable and correct. However, more expenses and time is required to contact the respondents.
(b) Questionnaires sent through Mail
In this method, also known as mail inquiry, the investigator prepares a questionnaire and
sends it by mail to the respondents. The respondents are requested to complete the
questionnaires and return them to the investigator by a specified date. The method is suitable
where respondents are spread over a wide area. Though the method is less expensive,
normally it has a poor response rate. Usually, the response rate in mail surveys has been
found to be about 40 per cent. The other problem with this method is that it can be adopted
only where the respondents are literate and can understand the questions. They should also be
able to send back their responses in writing. The success of the method depends on the skill
with which the questionnaire is drafted, and the extent to which willing cooperation of the
respondents is secured. For rural areas, this method has got its obvious limitation and is
seldom used.
(c) Interview by Enumerators
This method involves the appointment of enumerators by the surveying agency. Enumerators
go to the respondents, ask them the questions contained in the schedule, and then fill up the
responses in the schedule themselves. For example, this method is used in collecting
information during population census. For success of this method, the enumerators should be
given proper training for soliciting co-operation of the respondents. The enumerators should
be asked to carry with them their identity cards, so that the respondents are satisfied of their
authenticity. They should also be instructed to be patient, polite, and tactful. This method can
be usefully employed where the respondents to be covered are illiterate.
(d)Telephone Interview
In case the respondents in the population to be covered can be approached by phone, their
responses to various questions, included in the schedule, can be obtained over phone. If long
distance calls are not involved and only local calls are to be made, this mode of collecting
data may also prove quite economical. It is, however, desirable that interviews conducted
over the phone are kept short so as to maintain the interest of the respondent.
I.28
(iv) Questionnaire vs. Schedule
In the questionnaire approach, the informants or respondents are asked pre-specified
questions and their replies to these questions are recorded by themselves or by investigators.
In this case, the investigator is not supposed to influence the respondents. This approach is
widely used in main enquiries. In the schedule approach, the exact form of the questions to be
asked are not given and the task of questioning and soliciting information is left to the
investigator, who backed by the training and instructions has to use his ingenuity in
explaining the concepts and definitions to the informant for obtaining reliable information.
While planning a survey, preparation of questionnaire or schedules with suitable instructions
needs to be given careful consideration. Respondent’s bias and Investigator’s bias are likely
to be different in the two methods. Simple, unambiguous suitable wordings as well as proper
sequence of questions are some considerations which contribute substantially towards
reducing the respondents’ bias. Proper training, skill of the Investigators, suitable
instructions and motivation of investigators contribute towards reducing Investigator’s bias.
(v) Survey, Reference and Reporting Periods
Another aspect requiring special attention is the determination of survey period, reference
period and reporting periods.
i. Survey Period: The time period during which the required data is collected.
ii. Reference Period: The time period to which the collective data for all the units
should refer.
iii. Reporting Period: The time period for which the required statistical information
is collected for a unit at a time (reporting period is a part or whole of the reference
period).
The reporting period should be decided after conducting suitable studies to examine recall
errors and other non-sampling errors. For items of information subject to seasonal
fluctuations, it is desirable to have one complete year as the survey and reference period, the
data being collected every month or season with suitable reporting periods for the same or
different sets of sample units.
(vi) Sampling Frames
One of the main requirements for efficiently designing sample survey is a well constructed
sampling frame. In actual practice, quite often frames are not always perfect Various types of
imperfection such as omission, duplication etc. exists in the available frame. In multi-stage
sampling, the problems of securing a good sampling frame arise for each of the stages.
Usually a frame for higher stage units, such as towns, urban blocks and villages is more
stable than one for lower stage units such as farms and households, which are more subject to
changes. In agricultural surveys, normally the frames of first few stages of units up to village
level are used from records while the frame of households, fields etc. within the villages are
prepared afresh. This approach reduces the chances of imperfection in sampling frames.
(vii) Choice of Sampling Design
The choice of a suitable sampling design for a given survey situation is one of the most
important step in the process of planning sample surveys. The principle generally adopted in
the choice of a design is either reduction of overall cost for a pre-specified permissible error
or reduction of margin of error of the estimates for given fixed cost. Generally a stratified
uni-stage or multi-stage design is adopted for large scale surveys. For efficient planning,
various auxiliary information which is normally available is utilized at various stages e.g. the
I.29
area under particular crop as available for previous years is normally used for size
stratification of villages. If the information is available for each and every unit of the
population and there is wide variability in the information then it may be used for selecting
the sample through probability proportional to size methods The choice of sample units,
method of selecting sample and determination of sample size are some of the important
aspects in the choice of proper sample design.
(viii) Pilot Surveys
Where some prior information about the nature of population under study, and the operational
and cost aspects of data collection and analysis is not available from past surveys. It is
desirable to design and carry out a pilot survey. It will be useful for
i. testing out provisional schedules and related instructions,
ii. evolving suitable procedure for field and tabulation work, and
iii. training field and tabulation staff.
(ix) Field Work
While planning the field work of the survey, a careful consideration is needed regarding
choice of the field agency. For ad-hoc surveys, one may plan for ad-hoc staff but if survey is
going to be a regular activity, the field agency should also be on a regular basis. Normally
for regular surveys, the available field agency is utilized. A regular plan of work by the
enumerators along with proper supervision is an important consideration for getting a good
quality of data.
(x) Processing of Survey Data
The analysis of data collected in a survey has broadly two facets:
i. tabulation and summary of data and
ii. subject analysis.
The first task which is of primary importance is the reduction of collected data into
meaningful tables. The tables should be presented along with the background information
such as the objective(s) of the survey, the sampling design adopted, method used for data
collection and tabulation, and margin of error applicable to the results. These margins of
error provide the idea about the precision of estimates.
Subject analysis to be taken up after preparing summary tables, should include cross
tabulation of data by the meaningful, geographical, economy, demographic or other
breakdowns to study their relationship and trends among various characteristics. This is a
detailed technical analysis and is likely to be time consuming. Hence, this part should not be
tied up with the first part as otherwise the publication of the survey results might get delayed.
(xi) Preparation of Report
Although there are no set guidelines for presentation of results and preparation of report,
however some points which serve as guidelines in the preparation of sample survey reports
are given below:
i. Introduction & review of literature
ii. Objective(s)
iii. Scope
iv. Subject coverage
v. Method of data collection
I.30
vi. Survey references and recording
vii. Sampling design and estimation procedure
viii. Tabulation procedure
ix. Presentation of results
x. Activity of results
xi. Cost structure of the survey
xii. Agency for conducting the survey
xiii. References
2. Questionnaire Designing
Questionnaires and schedules are forms for recording the information as envisaged under the
survey. Designing of these is one of the most important aspects of the survey. The words
‘questionnaire’ and ‘schedule’ as per the current practice are generally used synonymously.
However, a technical distinction is sometimes made. The term questionnaire applies to forms
distributed through mails or given to informants to be filled in, by and large, without the
assistance or supervision of the interviewer, while a schedule is the form carried and filled in
by the investigator or filled in his presence.
The question as to whether the questionnaire or schedule approach is to be used in a survey
for collecting the required information needs consideration. In the former approach the
respondents are asked pre-specified questions and their replies to these questions are recorded
by themselves or by the investigators. This approach presumes that the respondents are
capable of understanding and answering the questions, since in this case the investigator is
not supposed to influence the responses in any way by his interpretation of the terms used in
the form. This method is widely used in mail inquiries. In the schedule approach, the exact
form of the questions to be asked are not given and the task of questioning and eliciting
information is left to the investigator, who backed by his training, experience and instructions
has to use his ingenuity in explaining the concepts and definitions to the informants for
obtaining reliable information. Detailed instructions are, however, given to the investigator
about concepts, definitions and procedures to be used in collecting data for the survey. In
various socio-economic surveys, the method of collecting data after meeting the respondents
and obtaining information of various characters by inquiry is commonly used.
From the above, it may appear that the schedule approach is subject to more investigator bias
than the questionnaire approach, as there is added scope in it for the investigator to influence
the responses of the informants. This will not be so, if well-trained and skilled investigators
are employed for the purpose. On the other hand the respondent bias may be substantial in
questionnaire approach, if the survey items are complicated and involve conceptual
difficulties. In such a situation, it would be desirable to train investigators for explaining the
terms involved rather than to burden the respondent with elaborate instructions and
clarifications. As the cost of questionnaire approach is generally less than that of schedule
approach, a decision as to which of the two methods should be followed in a particular survey
needs to be arrived at after carefully examining the possible effects of investigator and
respondent biases and the cost involved.
Designing of schedules/questionnaires with suitable instructions needs to be given careful
consideration in planning a survey as utility of the results of the survey depends to a large
extent on this. The framing of schedules or items should be done in a simple, unambiguous,
interesting and tactful manner and they should be so worded as not to influence the answers
of the respondents. The sequence of items is equally important. Those likely to help the
I.31
investigator in establishing a good rapport with the respondents should be put first and item
relating to a particular aspect of the survey should come together in a schedule/questionnaire.
As far as possible the items should be such that the answers can be recorded in numbers or
specific codes.
To reduce the non-sampling errors arising from ambiguous definitions and misunderstanding
of the questions by investigators/respondents, it is desirable to give some typical examples,
detailed explanatory notes and instructions for the items of information included in the
schedule/questionnaire. Clarification of doubts raised by the investigators is to be done in
such a manner that there is uniformity in the procedures followed by different investigators.
From what has been discussed above, it will appear that there are several considerations
which have to be kept in mind while designing the schedules. It is difficult to list out all of
them. There may be some which are specific to a particular survey and may require special
consideration. In the following paragraphs the main important considerations which should
be borne in mind while designing the schedule/questionnaire are given.
2.1 Three Kinds of Schedule Items
The information included in the schedule may be classified under the following three
headings:
2.1.1 Identification Information
This ensures that the schedule will not be misplaced or mixed-up, lost or duplicated; that the
information on it pertains to the particular sample case, and the interviewer and respondent
can be identified e.g. year, season, crop, name of the district, block, village, name of
cultivator and his father’s name etc. are entered against identification particulars.
2.1.2 Social Background or Census Type Factual Data
This information about respondent provides the variables by which the survey data are to be
classified and also the basis for evaluating the sample viz. cultivator’s total holding and
holding size group, category namely, SC, ST, or General, monthly income, total number of
family members, tenancy status, educational qualifications etc.
2.1.3 Questions on the Subject of the Survey
These questions may be directed towards obtaining more or less objective facts or towards
revealing attitudes and opinions on matters of current interest.
2.2 Considerations to be borne in Mind while Designing Schedules/Questionnaires
The first step in designing a schedule/questionnaire is to define the problem to be tackled by
the survey and hence to decide on what questions to be asked. The temptation is always to
cover too much, to ask everything that might turn out to be interesting. This must be resisted.
Lengthy questionnaires are as demoralizing for the interviewer as for the respondent, and the
questionnaire should be no longer than is absolutely necessary for the purpose.
2.3 Agency which will Make the Entries in the Schedules
If a highly trained investigator is to ask the questions and enter the replies, the form should be
different from the one drawn for informant to fill out himself since the interviewer can be
instructed regarding details which will ensure uniform definitions, entries and interpretations.
The terminology and questions should be adapted to the type of people who will give the
information. For example, a questionnaire addressed to specialist familiar with the subject
matter of the survey can be much more technical than the one directed to a cross-section of
I.32
the population. In designing schedules that are to be filled up by farmers, housewives,
employers etc. , the level of education should be taken into consideration.
2.4 Physical Appearance of the Schedule and Cooperation Received for the Survey
In surveys by mail, there is no doubt that an attractive looking questionnaire is a selling point
for cooperation. Consequently, an unattractive one may cause the recipient to put it aside or
even throw it. The fact that the form looks 'short', however, often contributes to securing
individual's consent to be interviewed. Informants will tolerate a short interruption of only to
get rid of the interviewer, but they may flatly refuse to answer a long list of questions.
2.5 How are the Questions to be worded?
The choice of the language used in expressing a question is of the greatest importance. It is
too often presumed that the respondents must be aware of the concepts and definitions used
in the questionnaire since these are obvious to the survey team. If the terminology is
ambiguous, the respondents will have to use their own judgment and different persons will
judge differently. This causes confusion and errors. Ambiguity arises with double barreled
questions, such as, the following question to a public transport "Do you like travelling on
trains and buses”? Respondent liking one and disliking other would be in a dilemma in
answering this question. Clearly it needs to be divided into two questions.
2.5.1 Use simple words which are familiar to all potential informants
The basic principle in good question wording is to use the simplest words that will convey the
exact meaning. Meaning of the questions becomes clear when the words used are well known
and mean the same thing to everyone. The question 'Do you operate land?' used in
agricultural surveys is poor. It is not clear whether the person is an owner cultivator or a
tenant cultivator.
2.5.2 Make the Questions as Concise as Possible
A question that contains long dependent or conditional clauses may confuse the informants.
In trying to comprehend the question as a whole he may over-look or forget clause and hence
his answer may be wrong. However, in opinion or attitude survey, it may be important to
have the complete question printed on the schedule.
2.5.3 Formulate the Question to Yield Exactly the Information Desired
The question should be self-explanatory. If the questions call for an answer in terms of units,
these units must be clearly defined. Suppose we want to ask the cultivator the seed rate used
for a specific crop. We should clearly mention whether the seed rate is to be reported in kg/ha
or in kg for the entire field.
2.5.4 Avoid Multiple Meaning Questions
Unless each question covers only one point, there will be confusion as to which one, the
answer applies to. Such items should be formulated as two or more questions so that separate
answer can be secured.
2.5.5 Avoid Ambiguous Questions
A question which means different things to different people is ambiguous. The best course is
to pre-test the questions through a pilot survey and thus detect ambiguities e.g. in a survey on
consumption of milk and ghee, suppose the question is about the quantity of milk and ghee
consumed during the month. It should be clearly mentioned whether it is during last calendar
month or one month prior to Investigator's visit.
I.33
2.5.6 Avoid Leading Questions
A leading question is one which, by its content, structure or wording leads the respondent in
the direction of a certain answer. In other words, all questions which produce biased answers
may be regarded as leading questions. Such questions should be avoided.
2.5.7 Keep to a Minimum the Amount of Writing Required on the Schedule
When feasible, use symbols for the replies. Explain these symbols somewhere on the
schedule. If the possible responses can be foreseen by pre-testing, the questions can be
answered as Yes or No, by writing a number, by putting a cross, by putting a symbol or by
encircling the correct answer.
2.5.8 Include a Few Questions that will serve as Checks on the Accuracy and
Consistency of the Questions as a Whole
Two questions that bring out the same facts though worded differently and placed in different
sections of the schedule, serve to check the internal consistency of the replies, e.g., in a socio-
economic survey, suppose we are asking the total holding of the farmer. It would be better if
we include the area owned, leased-in and leased-out separately in some block of the
proforma. This serves as a check to tally the total holding size.
2.6 Handbook of Instructions for the Field Staff
It would be desirable to prepare a comprehensive handbook of instructions explaining
concepts and definitions of various items for filling in the questionnaire/schedule under the
survey and a copy of the same should be supplied to each Field Investigator.
2.7 Sequence of Questions
Careful consideration should be given to the problem of the order in which questions should
appear. In order to guard against confusion and misunderstanding, questions should be
arranged logically, one question leading to the next. Specific questions should always follow
general questions. The opening question should be very interesting, this will ensure that the
respondents cooperate in parting with the desired information for the survey. Questions
which might embarrass the respondents should be placed towards the middle or end of the
questionnaires. Questions with an emotional tinge may be interspersed between items which
elicit more neutral reactions.
3. Conclusion
The problem of designing of questionnaires/schedules is not an easy task. Even if one
follows all the accepted principles, there usually remains a choice of several question forms,
each of which seems satisfactory. Every surveyor tries to phrase his questions in simple,
everyday language, to avoid vagueness and ambiguity and to use neutral wording. His
difficulty lies in judging whether, with any particular question, he has succeeded in these
aims. He may appreciate perfectly that leading questions are to be avoided but how can he
know which words will be 'leading' with the particular question, survey and population that
confront him, perhaps for the first time?
The answer to this question lies in detailed pre-tests and pilot studies, more than anything
else, they are the essence of a good questionnaire. However the experienced questionnaire
designer is, any attempt to shortcut these preparatory stages will seriously jeopardize the
quality of the questionnaire; past experience is a considerable asset, but in a fresh survey,
there are always new aspects which may perhaps not be immediately recognized, but which
exist and must be investigated through pre-tests and pilot studies.
I.34
References
1. Cochran, W.G. (1977). Sampling Techniques. Third Edition. John Wiley and Sons.
2. Des Raj (1968). Sampling Theory. TATA McGRAW-HILL Publishing Co. Ltd.
3. Des Raj and Chandok, P. (1998): Sample Survey Theory. Narosa Publishing House.
4. Murthy, M.N. (1977). Sampling Theory and Methods. Statistical Publishing Society,
Calcutta.
5. Singh, D. and Chaudhary, F.S. (1986). Theory and Analysis of Sample Survey
Designs. Wiley Eastern Limited.
6. Singh, D., Singh, P. and Kumar, P. (1978). Handbook of Sampling Methods.
I.A.S.R.I., New Delhi.
7. Singh, R. and Mangat, N.S. (1996). Elements of Survey Sampling, Kluwer Academic
Publishers.
8. Sukhatme, P.V. and Sukhatme, B.V. (1970). Sampling Theory of Surveys with
Applications. Second Edition. Iowa State University Press, USA.
9. Sukhatme, P. V., Sukhatme, B.V., Sukhatme, S. and Asok, C. (1984). Sampling
Theory of Surveys with Applications. Third Revised Edition, Iowa State University
Press, USA.
I.35
CROP ESTIMATION SURVEYS
U.C. Sud
I.A.S.R.I., New Delhi -110012
1. Introduction
Agriculture plays a very vital role in the Indian Economy. Large percentage of population in
our country is dependent on Agriculture. In view of importance of agriculture in our country,
collection and maintenance of Agricultural statistics becomes very important. India has very
well established agricultural statistics system. The system is essentially a decentralised one
with the State Agricultural Statistics Authorities (SASAs) involved in collection and
compilation of Agricultural Statistics at the state level while the Directorate of Economics
and Statistics in the Ministry of Agriculture (DESMOA) is involved in compilation of
Statistics at the National level. The system of data collection is quite comprehensive i.e.
estimates of various parameters of interest at the macro level are available with regard to crop
area and production, land use, irrigation, land holdings, agricultural prices and market
intelligence, livestock, fisheries and forestry. The agricultural statistics system has been
subject to review number of times so that it can adapt itself to the changing needs. Some of
the important expert groups which examined the working are: the Technical Committee on
Coordination of Agricultural Statistics (1949), the National Commission on Agriculture
(1976), the High Level Evaluation Committee (1983) and the Workshop on Modernization of
the Statistical System (1998). The National Commission on Agriculture (1976) made far
reaching recommendations which laid a strong foundation of the Agricultural Statistics
system. The High Level Evaluation Committee (1983) brought to light important data gaps
including methodological gaps. It identified newly emerging areas such as crop estimates at
the local level and crop forecasting. The Workshop on Modernization of the Statistical
System in India (1998) considered various measures which are required to modernize the
system.
The system of generating annual estimates of area, yield and production of crops in India is
more than a century old. A constant evaluation of the mechanism for generation of timely and
reliable agricultural production statistics, therefore, assumes vital importance and
significance. The Directorate of Economics and Statistics (DES) releases estimates of area,
production and yield in respect of 51 principal crops of foodgrains, oilseeds, sugarcane, fibres
and important commercial and horticulture crops. These crops together account for nearly
87% of agriculture output. The estimates of crop production are obtained by multiplication of
area estimates by corresponding yield estimates. Therefore, the estimates of area and yield
rates assume prime importance in the entire gamut of agricultural statistics. These are, along
with problems of implementation, being dealt with in Sections 2-8. The solution to the
problem of small area estimation of crop yield has been dealt with in Section 9.
2. Area Statistics
From the point of view of collection of crop area statistics, the States in the country are
divided into three broad categories:
i) The first category covers States and UTs which have been cadastrally surveyed and where
area and land use statistics are built up as a part of the land records maintained by the
revenue agencies (referred to as “Land Record States” or temporarily settled states). The
system of land records is being followed in 17 major states of Andhra Pradesh, Assam
(excluding hilly districts), Bihar, Chhattisgarh, Gujarat, Haryana, Himachal Pradesh,
Jammu & Kashmir, Jharkhand, Karnataka, Madhya Pradesh, Maharashtra, Punjab,
Rajasthan, Tamil Nadu, Uttar Pradesh and Uttaranchal and 4 UTs of Chandigarh, Delhi,
Dadra & Nagar Haveli and Pondicherry. These states/UTs account for about 86% of
reporting area. These states/UTs are covered under Timely Reporting Scheme (TRS) under
which 20% villages are selected at random in each season for complete area enumeration.
ii) The second category covers States where area statistics are collected on the basis of sample
surveys. A scheme for Establishment of an Agency for Reporting of Agricultural Statistics
(EARAS) has been introduced in these states, viz., Kerala, Orissa and West Bengal, and
later extended to Arunachal Pradesh, Nagaland, Sikkim and Tripura. The scheme
envisages, inter-alia, estimation of area through sample surveys in a sufficiently large
sample of 20% villages/investigator zones. These states account for about 9% of reporting
area.
iii) In the hilly districts of Assam, rest of the states in North-Eastern Region, (Other than
Arunachal Pradesh, Nagaland, Tripura and Sikkim), Goa, UTs of Andaman & Nicobar
Islands, Daman & Diu and Lakshwadeep where no reporting agency had been functioning,
the work of collection of Agricultural Statistics is entrusted with the village headmen. The
area statistics in these states are based on impressionistic approach. These states account
for 5% of the reporting area.
3. Yield Estimates
The second most important component of production statistics is yield rates. The yield
estimates of major crops are obtained through analysis of Crop Cutting Experiments (CCE)
conducted under scientifically designed General Crop Estimation Surveys (GCES). At
present over 95% of the production of foodgrains is estimated on the basis of yield rates
obtained from the CCEs.
Field Operations Divisions (FOD) of the National Sample Survey Office (NSSO) has been
providing technical guidance to the States and Union territories for organizing and
conducting Crop Estimation Surveys for estimating yield rates of principal crops. In addition,
NSSO in collaboration with States/Union Territories implements sample check programmes
on area enumeration work, area aggregation and conduct of crop cutting experiments under
the Scheme for Improvement of Crop Statistics (ICS). While executing the programme of
sample checks on crop cutting experiments, the FOD associates itself with the operational
aspects of the conduct of crop cutting experiments right from selection of sample villages,
training of field staff and supervision of field work and in the process, gathers micro level
information relating to conduct of crop cutting experiments for estimation of crop yield.
The results of Crop Estimation Surveys are analyzed and annual publication entitled
“Consolidated Results of Crop Estimation Surveys on Principal Crops” is brought out by the
NSSO regularly. The primary objective of GCES is to obtain fairly reliable estimates of
average yield of principal food and non-food crops for States and UTs which are important
from the point of view of crop production. The estimates of yield rates thus arrived at are
generally adopted for the purposes of planning, policy formulation and implementation. The
CCEs consist of identification and marking of experimental plots of a specified size and
shape in a selected field on the principle of random sampling, threshing the produce and
recording of the harvested produce for determining the percentage recovery of dry grains or
the marketable form of the produce.
3.1 Coverage
The crop-wise details of number of experiments planned under GCES during 2003-04 are given
in the following table:
I.37
Table 1: Number of experiments planned under GCES during 2003-04
Crops type/Crop
1. Food Crops Kharif Rabi Total
Paddy 117622 25667 143289
Wheat 0 70493 70493
Jowar 17676 12366 30042
Bajra 15676 0 15676
Maize 21986 2070 24056
Ragi 7050 0 7050
Barley 0 3450 3450
Gram 0 20435 20435
Sugarcane 23088 0 23088
Total 203098 134481 337579
2. Non-Food Crops
Groundnut 20265 5927 26192
Sesamum 9979 4637 14616
Castor 1912 0 1912
Rapeseed & Mustard 0 21503 21503
Linseed 0 5593 5593
Cotton 19768 0 19768
Jute 4420 0 4420
Mesta 849 0 849
Total 57193 37660 94853
Total (1+2) 260291 172141 432432
Other Food crops 71027 37246 108273
Other Non-Food Crops 12216 8992 21208
Total all crops 343534 218379 561913
Total no. of CCE in 2003-04 561913
Source: FOD, NSSO
4. Sampling Design
Stratified multi-stage random sampling design is generally adopted for carrying out GCES
with tehsils/taluks/revenue inspector circles/CD blocks/anchals etc. as strata, revenue villages
within a stratum as first stage unit of sampling, survey numbers/ fields within each selected
village as sampling unit at the second stage and experimental plot of a specified shape and
size as the ultimate unit of sampling as depicted in Figure 1.
I.38
Tehsil/Taluk
Revenue Village
Survey Number/Field
Experimental Plot
(Specified size/shape)
I.39
major reservoirs, availability/supply of important inputs including credit to farmers, rainfall,
and temperature.
5.2 Second Advance Estimates
The second advance estimate is made sometimes in the month of January every year when
the advance estimates of kharif crops prepared during the National Conference of Agriculture
for Rabi Campaign may undergo a revision in the light of flow of more precise information
from states. Around this time, the first advance estimates of rabi crops are also prepared. The
Second Advance Estimates then cover the second assessment in respect of Kharif Crops and
the first assessment in respect of Rabi Crops.
5.3 Third Advance Estimates
The third advance estimates are prepared towards the end of March/beginning of April every
year when the National Conference on Agriculture for kharif campaign is convened and the
states come up with their assessments for both kharif and rabi crops. The earlier advance
estimates of both kharif and rabi seasons are firmed up/validated with the information
available with State Agricultural Statistical Authorities (SASAs), remote sensing data as well
as the proceedings of CWWG.
5.4 Fourth Advance Estimates
The fourth advance estimates are prepared in the month of June/July every year when the
National Workshop on Improvement of Agricultural Statistics is held. Since most of the rabi
crops get harvested by the end of May, SASAs are in a position to supply the estimates of
both kharif and rabi seasons as well as likely assessment of summer crops during the National
Workshop.
Like third advance estimates, the fourth advance estimates are duly validated with the
information available from other sources.
5.5 Final Crop Estimates
Under the existing system of crop estimation, the fourth advance estimate is followed by final
estimates in December/January of the following agricultural year.
The main factors contributing to the relatively large number of crop estimates are the large
variations in crop seasons across the country and the resulting delay in the compilation of
yield estimates based on crop cutting experiments. As agriculture is a state subject, Central
Government depends upon State Governments for accuracy of these estimates. For this
purpose, State Governments have set up High Level Coordination Committees (HLCC), inter
alia, comprising senior officers from their Departments of Agriculture, Directorate of
Economics & Statistics, Land Records, and NSSO (FOD), IASRI, DES from Central
Government for monitoring and guidance in preparation of these estimates in a timely and
orderly manner.
In a bid to improve efficiency of estimates, pre-stratification of design taking into account the
impact of irrigation and type of seeds was adopted during 2004-05 in some states namely
A.P., Bihar, Chhattisgarh, Gujarat, Jharkhand, Karnataka, M.P., Maharashtra, Rajasthan and
Tamil Nadu. The details of such stratification are given in Table 2.
I.40
Table 2: Stratification according to inputs
State Crop for which pre-stratification is done according to
I.41
(ii) Stratification according to inputs may require conduct of larger number of
experiments (CCE) to draw statistically acceptable estimates of desired parameters.
Given the resources, it does not appear feasible to increase number of CCE at present.
6. Degree of Precision
The magnitude of standard error reflects the precision of the estimates and the degree of
precision reduced with increase in the standard error. It is generally agreed that desirable
level of standard error (SE) for crop yields is 0% to 5% . However, the experience shows that
in good number of cases, S.Es are above the desirable limits in some of states like Andhra
Pradesh, Bihar, Chhattisgarh, Gujarat, Karnataka, Kerala, Madhya Pradesh, Maharashtra,
Orissa, Rajasthan and West Bengal which put a question mark on the reliability of the
estimates. Concerted efforts are required on the part of State Governments by increasing
sample size and supervision to ensure that the SEs do not exceed 5%.
7. Limitations of CES
CES have been quite useful in providing desired estimates. However, it has the following
important limitations:
(i) Non response
(ii) Errors in CCE
(iii) Substitution of experiments
(iv) Delegation to Junior Officials
(v) Non availability of suitable equipments
(vi) Non response
The reliability of estimates depends to a great extent on the response level in conducting
CCE. In 2004-05, the overall response recorded was 94 percent as in the previous year.
Percentage response for food crops during Kharif and Rabi 2004-05 was 95% & 94%
respectively and for non-food crops, it was 96% & 92% respectively. Response was below 90
percent in Himachal Pradesh (86%) and Delhi (79%). Special measures are needed to
improve the response to more than 90 percent in these States to increase the acceptability and
reliability of estimates based on CCE.
7.1 Errors in CCE
An analysis of the results of CCE through the sample check under ICS few years before
revealed that about 90% of experiments could be conducted without error at All India level.
However, the position is quite different once State-wise analysis is made. Table 3 indicates
the position of different types of errors observed during the conduct of CCE in kharif season.
I.42
Table 3: Incidence of errors in Crop Cutting Experiments
Sl. States % of expts. % of expts. where error was noticed
No. where no
error was E1 E2 E3 E4 E5 E6 E7 E8
noticed
1 2 3 4 5 6 7 8 9 10 11
1 Andhra Pradesh 85 0 0 1 0 1 7 4 5
2 Assam 94 0 0 0 0 0 0 0 6
3 Bihar 92 0 0 0 0 0 4 0 4
4 Gujarat 92 0 0 0 0 0 9 0 0
5 Haryana 90 2 1 1 5 0 17 2 3
6 Himachal Pradesh 68 3 1 0 4 2 14 3 10
7 Jammu &Kashmir 75 0 0 4 0 0 15 4 3
8 Karnataka 79 1 2 0 11 1 1 5 8
9 Kerala 99 0 0 0 0 0 1 0 1
10 Madhya Pradesh 90 0 0 0 0 1 2 4 2
11 Maharashtra 44 5 4 2 36 4 37 15 15
12 Orissa 96 0 0 0 0 0 1 0 3
13 Punjab 92 0 0 1 0 0 4 0 5
14 Rajasthan 76 0 0 2 0 0 29 0 1
15 Tamil Nadu 75 0 0 3 0 0 28 0 2
16 U.P. 53 2 1 1 13 2 15 10 16
17 West Bengal 100 0 0 0 0 0 0 0 0
18 Delhi 55 0 0 0 0 0 20 0 35
19 Pondicherry 100 0 0 0 0 0 0 0 0
Note: E1: Error in selection of Survey/Sub survey nos., E2: Error in selection of field within
Survey/Sub-survey No., E3: Error in the measurement of the field, E4: Error in
selection of random nos., location and marking of plots, E5: Error in weighment of
produce, E6: Error in recording ancillary information, E7: Inadequate arrangements
for storing of produce for driage and incorrect reporting of constituents in mixture,
E8: Others
Mistakes that occur in the process of selection of survey numbers, selection of fields,
measurement of plots and selection of random numbers for location of plots have a direct
impact on the objectivity envisaged and results in upsetting the representative character of the
experiments. Instances of errors in reporting ancillary information provides adequate
evidence about the casual approach adopted by the primary worker even in eliciting the
I.43
relevant data. It is imperative to take strong administrative measures and impart intensive
training to field staff to ensure that prescribed procedure for conducting CCE is adhered to.
7.2 Substitution of Experiments
Instructions for conduct of CES prohibit substitution of sampling units once selected. But
instances of substitution of sampling units at village and field level are observed. During
1996-97, incidences relating to substitution were around 5% in Bihar, Karnataka and Madhya
Pradesh. In majority of the cases, the fields were substituted either because the crop was
harvested prior to selection of the field or crop harvested before the date fixed. Substitution of
duly selected sampling units by any other convenient unit may result in distorted results.
Better liaison between the primary workers and the cultivators would go a long way to
control such high incidence of substitution.
7.3 Delegation to Junior Officials
The field work of crop estimation surveys is entrusted with the officials who are normally
one rank higher than the primary worker for area enumeration. Delegation of crop cutting
work particularity to the junior rank has been observed in several states. For example, the
delegation of work was of the order of 10% in case of U.P. during summer 1996-97.
Adequate arrangements are needed for ensuring proper training to the field workers entrusted
with CES to avoid improper conduct of crop cutting experiments.
7.4 Non Availability of Suitable Equipments
While an untrained worker cannot conduct the experiment properly, supply of essential
equipments and its proper use is required for accuracy in measurements. The position is far
from satisfactory particularly in the case of Bihar, Haryana, Himachal Pradesh, J&K,
Karnataka, Maharastra, Punjab, Rajasthan and U.P. Even the supplied equipments were
reported to have not been carried to the field for the conduct of CCE in many cases in
Karnataka, M.P., Maharashtra, Rajasthan and U.P. This calls for strong administrative
measures for effecting further improvement. Table 4 gives the percentage of experiments
conducted without use/improper use of crop cutting equipments during 5 years preceding
1997-98 as observed through ICS.
Table 4: Supply and use of equipments for CCE during 1993-1998
1 2 3 4 5 6 7 8 9
1997-98 15 16 36 39 20 18 26 24
1996-97 18 61 37 40 18 17 25 24
1995-96 17 60 37 39 18 16 24 23
1994-95 16 60 34 39 18 17 23 22
1993-94 16 60 37 38 18 18 24 23
I.44
8. Schemes for Fine Tunning of Crop Statistics
Availability of reliable and timely estimates of area and production assumes prime
importance as these estimates are used by the government for taking a number of policy
decisions regarding production, pricing, processing, procurement, storage, transport, export-
import, public distribution etc. In its quest for improving quality, reliability and timeliness of
agricultural statistics, DES has initiated the following important schemes:
(i) TRS
(ii) ICS
(iii) EARAS
(iv) FASAL
8.1 Timely Reporting Scheme (TRS)
The agricultural statistics has come under severe criticism within and outside the Government
for considerable time-lag in the availability of desired statistics. To positively respond to this,
DES initiated TRS in 1969-70 in the land record States to reduce time lag between the period
of sowing and the availability of estimates of area sown on one hand and between completion
of harvesting and availability of estimates of production in respect of important crops on the
other hand. The primary objective of this scheme is to obtain reliable estimates of area and
production of principal crops in each season with break up of area under irrigated/ unirrigated
and traditional/high yielding varieties of crops on the basis of priorities enumeration
conducted in 20% of villages. These estimates are required to be furnished to the Govt. of
India by 30th November for kharif crops and by 30th April for rabi crops. Under the scheme,
Girdwari is conducted by revenue agency of the state on priority basis in a sample of 20%
villages every year. The enumeration of area done by Patwaris is supervised by trained staff
of NSSO in random sample of 20% villages in each season in a state. Thus, five randomly
selected non overlapping samples surveyed each year cover all the villages in five years. The
TRS results have been found extremely useful both by states and DES. In addition to timely
and reliable area estimates, TRS also serves as a sampling frame for CES and related aspects.
Thus, the TRS scheme has an important role in serving the following dimensions of
Agricultural Statistics System:
(i) Sample based reliability of crop area statistics taking into consideration the heavy work
load of State primary workers.
(ii) Timeliness in the aggregation and flow of the area estimates envisaging the area
enumeration on priority for the sampled villages.
(iii) Frame for the General Crop Enumeration Surveys facilitating the methodological
commitments of scientifically designed crop yield estimation exercise.
(iv) Frame for the Agricultural Census Surveys.
(v) Frame for Input Surveys.
8.2 The Scheme for the Improvement of Crop Statistics (ICS)
The scheme for Improvement of Crop Statistics was initiated in 1973-74 with the main
objective of locating through joint efforts of Central and State agencies from year to year, the
deficiencies in the State system of crop statistics and suggesting remedial measures to effect
lasting improvement in the system. The programme of ICS is in operation in 16 land record
States/UTs. and the 3 non-land record States of Kerala, Orissa and West Bengal. The scheme
provides for exercising checks at every stage of work relating to estimation of crop
I.45
production on an equal matching basis by the Central as well as State supervisory staff on
three basic aspects namely area enumeration, area aggregation and conduct of CCE. The
programme envisages the following:
(i) Studying the State system of crop estimation in its normal operative conditions.
(ii) Identifying the deficiencies and weaknesses in the system.
(iii) Physical verification of crop enumeration done by the Patwaris in a sample of 10,000
villages,
(iv) Checking the accuracy of the area statistics transmitted from the village level through
crop abstract in the very same 10,000 villages.
(v) Providing technical guidance and supervision at the harvest stage in the conduct of about
30,000 CCE distributed across principal crops in various states.
The above checks specifically relate to crop area enumeration, page totalling of khasra
register and supervision of harvest stage of CCE. The sample checks are undertaken by the
supervisory staff of the National Sample Survey Office and the staff of the State agencies on
a matching basis over a non-overlapping sample. The samples for this purpose are drawn
following the stratified multistage random sampling design. The analytical findings of the
ICS scheme bring to focus the precise lines along which the improvement in the system of
crop statistics system could be effected in context of the conditions prevailing in each State.
The survey operations for generating crop area and yield estimates are expanded vastly over
the space and involve multiple agencies at various levels. Though these surveys provide the
estimates of comparable nature, the management of the survey and control of survey
operations differ significantly from state to state.
The surveys of such magnitude face a potential threat to the quality of their output from
numerous sources of non-sampling errors. The scheme of ICS serves the purpose of an
objective assessment of non-sampling errors in these surveys and identifies their sources as
well as the possible impact on the quality of data.
8.3 Establishment of an Agency for Reporting Agricultural Statistics
In three “permanently settled states” namely Kerala, Orissa and West Bengal, there is no land
record system and there is no regular agency for collection of agricultural statistics. In order
to bridge data gap, a scheme namely Establishment of an Agency for Reporting Agricultural
Statistics (EARAS) is being implemented in States which do not have permanent land record
system. In the North-Eastern (NE) States (except Assam) where no reporting agency
functions, the land use statistics are generated through ad-hoc methods. EARAS scheme has
been extended to NE States of Arunachal Pradesh, Nagaland and Tripura besides Sikkim
during 1994-95 but data generation is yet to commence. In these states, the land use statistics
are still generated through ad-hoc methods. However, the scheme of EARAS has been of
immense significance in generating crop estimates for the “permanently settled states”. Under
the scheme, estimates of area and yield are built on the basis of complete enumeration of 20%
sample of villages every year. The enumeration is supervised by trained staff of NSSO. This
scheme also has some weaknesses as primary workers either do not complete work in time or
do not discharge duties with required sincerity which results in under estimation of area and
production of various crops.
I.46
8.4 Forecasting Agricultural Output Using Space, Agrometeorology and Land – Based
Observations (FASAL)
Timely availability of reliable information of agricultural output and other related aspects is
of great significance for planning and policy making. The existing system of agricultural
statistics, in spite of established procedures and wide coverage, has inherent limitations in the
matter of providing an objective assessment of crops at the pre-harvesting stages with the
desired spatial details, which are essential to identify problem areas and the nature of required
interventions in terms of spatial, temporal and qualitative inferences.
Capabilities of the existing system of crop forecasts and crop estimation can be enhanced
with the introduction of technological advancements and the adoption of emerging
methodologies. In turn, an efficient and sound information mechanism can assist
considerably in the management of concerns in areas such as food security, price stability,
international trade, etc. Remote Sensing (RS), Information and Communication Technology
(ICT) and Geographic Information System (GIS) can be used towards this end.
In the year 1987, the Department of Agriculture & Cooperation (DAC) sponsored a project
called "Crop Acreage and Production Estimates (CAPE)".The project is being implemented
through the Space Application Centre (SAC), Ahmedabad. The objective of CAPE is to use
remote sensing (RS) techniques for crop area and production forecasting. Under the CAPE
project, basic procedures, models, and software packages for crop area and production
forecasting, using remote sensing and weather data have been developed. The work is being
carried out jointly by Space Applications Centre (SAC), State Remote Sensing Agencies
(SRSA), State Departments of Agriculture, and State Agricultural Universities/Institutions.
As a result, considerable human resource development has taken place and initial expertise
generated. The methodology and tools developed under CAPE have been discussed in
different fora and found to have attained a reasonable degree of standardization for
integration into the main system.
CAPE project has successfully demonstrated national level forecast of wheat and kharif rice,
in addition to making district level pre-harvest production forecasting of cotton, sugarcane,
rapeseed/mustard and rabi sorghum in their major growing regions in the country using RS
technology and other auxiliary information. It has successfully overcome the problem of
persistent cloud cover during kharif season by using Synthetic Aperture Radar (SAR) data
from Radarsat. Moving a step further, the FASAL project aims at multiple in season
forecasting of crop area and production.
Crop forecasting at the planting stage can be made by employing econometric and agro-met
models using previous years crop acreage and production, market price, current season
weather forecast/data, and other auxiliary information. At this stage, farmer’s response may
be unpredictable. Remote Sensing (RS) technique becomes effective for crop assessment
once it attains sufficient ground coverage, say around 45 days after sowing. RS data provides
information regarding area covered by crop, its condition and the likely yield. This coupled
with agro-meteorological observations and limited field observations can provide fairly
accurate information on likely yield of the crop.
Remote Sensing, weather and field observations provide complementary and supplementary
information for making crop forecasts. Thus an approach, which integrates inputs from the
three types of observations, is needed to make forecasts of desired coverage, accuracy, and
timeliness. As such, this project has been named as “Forecasting Agricultural output using
Space, Agro-Meteorology and Land based Observations” or FASAL.
The FASAL project has been formulated with the following considerations:
I.47
(i) Experience of CAPE in implementation of RS (Remote Sensing)-based crop production
forecasting experiment,
(ii) The requirements of the MOA/DAC in terms of timeliness, accuracy and coverage of
crops,
(iii) Experience and recent studies outside India for RS-based crop assessment/ forecasting,
(iv) Improvements available in terms of RS technology as well as in related disciplines, and
(v) Use of other information such as weather based and field surveys for crop forecasting.
The concept of FASAL thus seeks to strengthen the current capabilities of early and in-season
crop estimation capabilities from econometric and weather based techniques with RS
applications. Mid-season assessments can be supplemented with multi-temporal coarse
resolution data based analysis. In the latter half of crop growth period, direct contribution of
RS in the form of acreage estimates and yield forecasts would be available.
However, in this case also, the addition of more extensive field information and weather
inputs would increase the forecast accuracy.
9. Small Area Estimation of Crop Yield
In recent years the thrust of planning process has shifted from macro to micro level. There is
demand by the administration and policy planners for accurate estimates of various parameter
at the micro level. In our country estimates of crop yield are, currently, available at some what
higher level i.e. district/block level. In what follows, we give below an approach of crop yield
estimation at the Gram Panchayat (GP) level. GP level yield estimation is very crucial in the
context of Rashtriya Krishi Bima Yojana (RKBY). The Bima Yojana is a scheme related to
compensation to be provided to the farmer in case he suffers a yield loss. The scheme involves
assessment of yield loss at the GP level for decision making on compensation to be provided
to the farmer.
In view of the demands of modern time the thrust of research effort has also shifted to
development of precise estimators for small areas. An off shoot of this development is that
various small estimation techniques are being proposed by the researchers for implementation.
The basic rationale of the approach of small area estimation is that certain assumption/models
are conceptualized which are assumed to hold good for small as well as large area. In the
context of Rashtriya Krishi Bima Yojana information on some auxiliary variable at the village
panchayat level needs to be generated to scale down the crop yield estimates obtained in the
present system of General Crop Estimation Surveys (GCES) at district/block level, for
developing the village panchayat level estimates. Past experience suggest that information
about crop area, crop yield as obtained through farmer’s appraisal, using well tested structured
questionnaires regarding the crop production obtained through random sampling approach has
been found to perform satisfactorily in many practical situations. Studies conducted in several
African countries (Verma, Merchant and Scott; long acre report, 1988) suggest that such
estimates as obtained through farmer by inquiry are in close agreement with the actual
production figures. Studies at IASRI also revealed a fairly high degree of correlation between
the farmer’s estimates and the estimates obtained through the crop cut approach. The farmer is
expected to provide reasonably accurate estimate of crop production if this data is obtained
within 2-3 days of harvest. It is apprehended that farmers’ appraisal is likely to be affected
qualitatively in view of interest of the stakeholders at village panchayat level. However, this
information, if used judiciously, only to workout (say) correction factors of farmer estimates,
which is then used for scaling down the estimates of crop yields of GCES series, the resulting
estimates are likely to be less influenced by the effect of stakeholders interest. Even though
I.48
the farmer’s estimates are likely to be influenced by their own interest, the proportions are
expected to be almost free from such effects provided the under estimation or over estimation
is behaving in a similar way in the entire area. This would be the case if every farmer is
insured and behaves as a stakeholder.
The approach is akin to small area approach in the sense that Block/District levels estimates
are scaled down to GP level estimates under the assumption that over estimation and under
estimation are behaving in similar way in the entire area. One of the main advantage of the
approach is that the official estimates at the higher level (Block/District) need not be
disturbed. Also the production estimates as obtained at GP levels are in conformity with Block
level estimates of crop production.
To the extent possible farmers’ inquiry based data should be collected by an agency which is
different than the one used for collecting the crop-cutting experiment data. Perhaps the village
based persons are best suited for collecting farmers’ inquiry data. This will considerably
reduce the data collection cost.
References
1. Cochran, W.G., (1939). The use of analysis of variance in enumeration by sampling. J. Amer.
Statist. Assoc., 34, 492-510.
2. Des Raj (1968). Sampling theory. Tata-Mcgraw-Hill Publishing Company Ltd.
3. Cochran, W.G., (1977). Sampling techniques. Wiley Eastern Ltd.
4. Hansen, M.H. and Hurwitz, W.H. (1943b). On the theory of sampling from finite populations.
Ann. Math. Statist., 14, 333-362.
5. Hansen, M.H., Hurwitz, W.H. and Madow, W.G., (1993). Sample survey methods and theory,
Vol. 1 and Vol. 2. John Wiley & Sons, Inc.
6. Kish, L. (1965). Survey sampling. John Wiley & Sons, Inc.
7. Lahiri, D.B. (1954a). Technical paper on some aspects of development of the sample design;
National Sample Survey Report No. 5, Government of India, reprinted in Sankhya, 14, 264-
316.
8. Murthy, M.N., (1977). Sampling theory and methods. Statistical Publishing Society.
9. Sukhatme, P.V., Sukhatme, B.V., Sukhatme, S. and Ashok, C. (1984). Sampling
theory of surveys with applications. Indian Society of Agricultural Statistics.
10. Yates, F. (1949). Sampling methods for Censuses and Surveys. Ist ed.; 2nd ed.
(1953); 3rd ed. (1960).
I.49
AGRICULTURE CENSUS
U.C. Sud
I.A.S.R.I., New Delhi -110012
1. Introduction
Agriculture Census is a large scale periodic, Government sponsored operation for collection
and derivation of quantitative information about nation’s agriculture, using ‘agricultural
operational holding’ as the statistical unit. The operational holding is defined as “all land
which is wholly or partly used for agriculture production, and is operated as one technical
unit by one person alone or with others without regard to title, legal form, size or location.” A
technical unit is defined as the unit which is under the same management and has the same
means of production such as labour force, machinery and animals etc. Agricultural
production includes growing of field crops, fruits, grapes, seeds, tree nurseries (except those
of forest trees), vegetables and flowers, production of coffee, tea, cocoa, rubber, jute,
oilseeds, fodder grass etc. Grass would be treated as a crop if special efforts are made to raise
it.
The concept of agricultural operational holdings adopted in India differs to some extent with
that of Food and Agricultural Organization (FAO), as it does not include those holdings
which are not operating any agricultural land and are engaged exclusively in livestock,
poultry and fishing etc.
2. Essential Features
The basic unit of information in the Agriculture Census is the operational holding as distinct
from the ownership holding. It is because of the fact that it is the operational holding which is
a fundamental unit of decision-making in programmes of agricultural development. The
Agriculture Census provides essential information not only for the country as a whole and for
the States but also for administrative units like district, tehsil and community development
blocks. The data in Agriculture Census is collected according to different size classes of
operational holdings, social groups like schedule castes, schedule tribes and others and
gender.
3. Brief History of Agriculture Census in India
Agriculture Census in India forms part of World Agriculture Census programme of FAO. The
first Agriculture Census in India was conducted in 1970-71. So far, eight Agriculture
Censuses with reference years 1970-71, 1976-77, 1980-81, 1985-86, 1990-91, 1995-96,
2000-01 and 2005-06 have been completed and the 9th Agriculture Census with reference
year 2010-11 is in progress. The reference period in Agriculture Census is the agriculture
year starting from July to June.
4. Frequency
The Agriculture Census in India is conducted on the interval of Five-Years i.e.
quinquennially differing from the World Census of Agriculture which is recommended to be
carried out decennially. The reason for conducting Agriculture Census in India
quinquennially is that the structure of holdings in this country changes frequently due to
fragmentation of holdings, urbanization and other factors, like, industrialisation.
5. Legal Framework
Though, agriculture is a State subject, the responsibility for conducting Agriculture Census
lies largely with the Centre as Agriculture Surveys are included in the concurrent List of
Indian Constitution.
6. Survey Reference and Reporting Periods
The reference period is the time period to which the data should refer. The survey period is
the time period during which the required data are collected. The Agriculture Census in India
is conducted with the participation of administrative machinery of States/UTs. In Land
Record states, which account for about 86% of the reported area, the data is collected from
the revenue records and in Non-land Record States the data is collected on sample basis. In
Land Record states, the fieldwork for the Agriculture Census starts only after the records for
Census reference year are updated which generally takes about 6-8 months after end of the
year. Thus, the actual data collection activity commences only 6-8 months after the end of
reference period. For example, for Agriculture Census 2005-06 (reference period: July 05
June 06), data collection started sometimes after January, 2007.
7. Instruction Manuals
Quality of data collected during a census/survey depends on the quality of field work
performed by the enumerators and the supervisors. Field staff must understand clearly all the
details and procedures to be followed and learn a large number of concepts and definitions. It
is almost impossible for them to become fully conversant with these during short training
period and they, therefore, need printed material for reference. The manuals will enable them
to review what they have been taught in order to master the subject matter and consult those
items where doubts or problems arise as they proceed with the data recording. The instruction
manuals fulfill two primary purposes. The first is to serve as an instrument of study during
training courses and the second to provide basic material for reference during enumeration.
With instruction manuals, it is much easier to achieve and maintain data comparability.
The instruction manuals are prepared well in advance of training and made available to states.
The language of the instruction manuals is kept clear and simple so that they are easily
understood. As much as possible, words which can be interpreted in many ways or words
which differ in meaning from one locality to another are avoided. In view of different system
of data collection in Land Record and Non-Land Record states, separate manuals are
prepared for both the categories. However, the manual for Input Survey is common to all the
states. The manuals are translated in local language before these are used for training and
distributed to field staff.
8. Training
Agriculture census collects data on a number of agricultural operations and the methodology
for obtaining such data may be different for different items. Thus taking an agricultural
census is very complex because of variability in the socio-economic structure between
provinces. Therefore, the training of census/survey staff is organized in a planned and
coordinated manner.
9. Methods of Data Collection
An agricultural census operation involves collection of data on a large number of items.
Most of the characteristics on which data are needed are of quantitative nature. In a number
of developed countries where holders are literate and maintain records self-enumeration is the
preferred approach. In most developing countries, however, holders do not have records of
their holdings. They may not even know in definable units of measurement the amount of
I.51
land they operate, particularly in areas where cadastral records do not exist. Collecting data
through an interview has many limitations under such circumstances. In India, the
Agriculture Census data is collected following two broad approaches-in Land Record States,
the information on number and area of operational holdings are collected by the village
accountant or Patwari from a register popularly known in northern India as Khasra. The
village accountant has been called by different names in different parts of the country such as
Karnam in South, Telathi in Maharashtra, Karamchari in Bihar, lekhpal in Uttar Pradesh etc.
This category covers around 86% of total reported area of the country.
The data collected by village accountant are on the basis of complete enumeration is called
girdawari. Most of the geographical areas of temporarily settled states are cadastrally
surveyed and detailed maps are available in tehsil and district Headquarters. The statistics
obtained by different village accountants are aggregated to get the statistics at higher
administrative units such as blocks, tehsils, district, states etc. This system of data collection
is being followed in 18 states namely, Andhra Pradesh, Assam (excluding hill districts).
Bihar, Chattisgarh, Gujarat, Haryana, Himachal Pradesh, Jammu & Kashmir, Jharkhand,
Karnataka, Madhya Pradesh, Maharashtra, Punjab, Rajasthan, Tamil Nadu, Puduccherry,
Uttaranchal and Uttar Pradesh and union territories Chandigarh, Dadra and Nagar Haveli,
Andman & Nicobar Islands and Delhi.
In states other than those mentioned above (non-land record states), the data is collected on
sample basis. A sample of 20% villages is drawn for estimating number and area operated,
farm characteristics and sample of 7% villages for Input Survey, though there are, at times,
some variations in some states. The 20% villages selected for the purpose are same as those
selected for Establishment of an Agency for Reporting Agricultural Statistics (EARAS).
It can thus be seen that for collection of data in Agriculture Census, an approach of census-
cum-sample survey has been adopted. It would be instructive to see the advantages and
disadvantages of carrying out Census on complete enumeration basis.
10. Advantages of Carrying Out Census on Complete Enumeration Basis
i. Results can be obtained for small administrative units. Such information is sometimes
required for local planning, for practical purposes like planning irrigation projects or for
re-organization on agro-climatic and ecological basis. Sample estimates become less
accurate for small administrative units since sample size becomes thin.
ii. Information on crops which are cultivated to a limited extent but are of great economic
importance or information on rare species of animal available only in a specific
geographical area can be obtained reliably.
iii. Census listings can be used as a frame for purposes of future sample surveys.
iv. Processing data from a complete enumeration is straight forward and does not involve
the calculation of sampling errors or expansion factors and thus requires less skill.
11. Disadvantages of Carrying out a Census on Complete Enumeration Basis
i. Conducting a census on complete enumeration basis in countries with large number of
agricultural holdings is more expensive and time consuming than a census conduced on
sample basis. This is a particularly important consideration in areas with poor
communication and transport facilities.
I.52
ii. For a complete enumeration census, very large number of enumerators and supervisors
are required. Often, required number of candidates with desired qualifications are not
available leading to lowering of standards with consequent effect on data quality.
iii. For a complete census, the quantity of data to be processed is very large. The results are
often delayed due to non-availability of sufficient data processing facilities. Also, the
cost of data processing is higher due to large volume of data involved.
12. Different Phases in Agriculture Census
i. Agriculture Census operation is carried out in three phases. During phase-I, a list of all
the holdings with data on area, gender and social groups of holder is prepared with the
help of listing schedule L-1/L-2. This operation covers all the villages in land record
States and 20% of the villages in non-land record States. During this phase, one more
schedule L-3 relating to village summary providing information on village infrastructure
is also filled up.
ii. During phase-II, detailed data on tenancy, land use, irrigation status, area under different
crops (irrigated and unirrigated) etc. are collected in holding schedule known as
“Schedule-H”.
iii. Phase-III, which is called as Input Survey, relates to collection of data on pattern of input
use across various crops, States and size groups of holdings, in addition to data on
agriculture credit, implement and machinery, livestock and seeds. So far, seven Input
Surveys with reference years 1976-77, 1981-82, 1986-87, 1991-92, 1996-97, 2001-02
and 2005-06 have been completed and the eighth one with reference year 2010-11 is
under progress. This survey is conducted in 7% of villages selected at stratum level. In
this survey, the second stage sampling is also adopted in which 20 holdings in all are
selected from each sampled villages out of the different size groups of holdings like
marginal (below 1.0 ha.), small (1.0 – 2.0 ha.), semi-medium (2.0 – 4.0 ha.), medium
(4.0 – 10.0 ha.) and large (10. 0 ha. and above). Four holdings are selected randomly for
collection of data from each such groups.
13. Processing of Survey Data
The rapid improvement in electronic data processing has greatly facilitated the task of data
processing. However, the proper choice of hardware - whether personal computers or main
frame computers-requires knowledgeable input. While considering the hardware
requirements, the main characteristics of census data processing which are kept in mind are:
i. Large amount of data to be entered in a short time.
ii. Large amount of data storage required
iii. Relatively simple transactions
iv. Relatively large number of tables to be printed
v. Extensive use of raw data files which need to be on-line, if possible.
Data entry, which refers to transfer of data from questionnaires to the computer-readable
media, is one of the greatest time and resource consuming phases of data processing.
Considerable time is required to write computer programmes for data entry, error
identification, automatic error correction (if applied), tabulation and calculation of sampling
error etc. The computer software is thoroughly tested before actual use. The main task of
software in census data processing is:
I.53
i. Data entry
ii. Data validation/cleaning
iii. Automatic data correction (when applied)
iv. Data tabulation
v. Presentation of data for printing
In addition, in case of sample enumeration, estimates have to be expanded and provision for
calculating expansion factors and sampling errors is required. The data processing work of
current Agriculture Census is being done by National Informatics Centre (NIC) under an
MOU signed between Department of Agriculture and Cooperation and the NIC.
14. Seventh Agriculture Census 2000-01
In the Agriculture Census 2000-01, the scope of earlier schedule-L3 was enhanced. This
schedule provides the information on Village Level Agricultural Infrastructure facilities
available in all the villages in the country. In this schedule the statistical unit for data
collection is village and not the operational holding. Information pertaining to the community
as a whole like community ponds in the village maintained by panchayats/local body for
fishing, number of godowns, cold storage, fertilizer/pesticide dealers, seed dealers,
agriculture credit institutions, veterinary centers, regulated marketing centers etc were
collected in this schedule. The information pertaining to the connectivity of the village by all
weather roads was also collected.
The items covered for data collection in Agriculture Census may be seen at Annexure-I.
To enhance the utility of Agriculture Census data, information on horticulture crops was also
collected.
15. Input Survey 2001-02
As mentioned above, the data in Input Survey is collected through household enquiry in the
7% selected villages from each stratum i.e. taluka. For the second stage sampling, four
operational holdings from each size groups i.e. marginal, small, semi-medium, medium and
large holdings are selected randomly. In this survey the sampling frame is the list of
operational holdings prepared at the time of phase-I operation. Since, Input Survey is
conducted after one year of the Agriculture Census, the list is updated by visiting each
household in the village and additions/deletion is done by the enumerators wherever
applicable. In this Survey, institutional holdings are not covered and only the resident
cultivators of the selected villages are considered for enquiry. Various additions in coverage
of items were done. The list of items on which data in this Survey was collected may be seen
at Annexure-II.
16. Dissemination of Data
Three repots are brought out containing the Agriculture Census data:-
I.54
For wider dissemination of the information to general public, policy planners, researcher etc,
the results of Agriculture Census and Input Survey have been put-up on the Department’s
website http://agcensus.nic.in. One can get the data of Agriculture Census with reference
years 1995-96 and 2000-01 and Input Survey with reference years 1996-97 and 2001-02 on
this website. Also, All India Reports on Agriculture Census and Input Survey, Schedules and
Manual of Instructions of Agriculture Census and Input Survey have also been put-up on the
web.
17. Errors
Considering the large number of enumerators and supervisors employed, the number of steps
involved in organization, difficulties in controlling operations, particularly in remote areas, it
is also necessary to check the quality of data disseminated before released for public use.
Data users should be aware of data limitations in order to avoid mistakes in decision making.
Annexure-I
all the tables give data by size group and social group. In addition, table-1 gives data on gender
and types of holdings, viz., individual, joint and institutional.
I.55
Annexure-II
TABLE GENERATED THROUGH INPUT SURVEY
I.56
Table Description User/Uses
No.
Phase-III (Input Survey)
(B), Agriculture and ICAR for
(C) estimating the demand supply
position of certified seeds
9D Methods of pest control Used by Ministry of
Agriculture and ICAR
10 Educational status of operational holders For knowing how educated and
technically qualified is
operational holder
It helps in formulating
programmes for targeted
groups considering their
educational status.
Table-1
Number of Holdings and Operated Area, India
Agriculture Census 2000-01
Table-2
Distribution of Number of Holdings and Area Operated in India, 2000-01
Size group Number Area Average Percentage Percentage of
of operated area per of area operated to
holdings (in holding holdings total area
(in million to total
million) hectares) holdings
Marginal 75.41 29.81 0.40 62.88 18.70
(Below 1.0 ha)
Small 22.69 32.14 1.42 18.92 20.16
(1.0 – 2.0)
Semi-medium 14.02 38.19 2.72 11.69 23.96
(2.0 – 4.0 ha.)
Medium 6.58 38.22 5.81 5.48 23.97
(4.0 – 10.0ha.)
Large 1.23 21.07 17.12 1.03 13.22
(10.0 ha. & Above)
All holdings 119.93 159.44 1.33 100.00 100.00
I.57
References
1. Colwell (1997). Looking into Agricultural Statistics: Experiences from Asia and the
Pacific, CGPRT Working Paper 29. Bogor
2. FAO (1967). Programme for the 1970 World Census of Agriculture. Rome
3. FAO (1989). Sampling methods for Agricultural Surveys. Rome
4. FAO (1995). Programme for the World Census of Agriculture 2000. Rome
5. FAO (1996a). Conducting Agricultural Censuses and Surveys. Rome
6. FAO (1997a). Guidelines on Employment: Supplement to the Programme for the
World Census of Agriculture 2000. Rome
7. Sukhatme, P.V. (1959). Major developments in the theory and application of sampling
during the last twenty five years. Estadistica, 17, 62-679.
8. Yates, F. (1949). Sampling methods for Censuses and Surveys. Ist ed.; 2nd ed. (1953);
3rd ed. (1960).
I.58
HORTICULTURE SURVEYS
Tauqueer Ahmad
I.A.S.R.I., New Delhi -110012
1. Introduction
In recent years, horticulture sector has emerged as an important component of the Indian
economy. This sector of agriculture contributes about one-fifth share in the economy of
agriculture and allied sectors. Hence, the statistics of horticultural crops has become one of the
priority programmes for the planning commission. Horticulture development is being given high
priority in the Five Year Plans. For preparation of various developmental programmes and for
policy formulations etc., the availability of adequate, reliable and timely statistics on area, yield
and production estimates of horticultural crops is essential. At present, a scheme namely "Crop
Estimation Survey on Fruits and Vegetables CES-F&V" is being implemented by Directorate of
Economics & Statistics (DES), Ministry of Agriculture, Govt. of India. This scheme has so far
been implemented only in 11 states namely Andhra Pradesh, Gujarat, Haryana, Himachal
Pradesh, Karnataka, Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu and Uttar Pradesh and
covers only 7 fruits and 7 vegetables namely, Mango, Apple, Banana, Grapes, Guava, Citrus,
Pineapple, Cauliflower, Potato, Onion, Tomato, Cabbage, Ginger and Turmeric. For estimation
of area and production of fruit and vegetable crops, methodology developed by Indian
Agricultural Statistics Research Institute (IASRI) is being used in these states. Here, present
method of generation of data, existing methodology being used and an alternative methodology
for estimation of area and production of different horticultural crops have been discussed in
brief.
IASRI carried a series of surveys to evolve a sampling methodology for estimating area and
yield of fruits and vegetables. A study on vegetables was conducted in rural areas of Delhi by
Sukhatme et al. (1969). Singh et al. (1976) conducted a study on fresh fruits in Tamil Nadu and
on vegetables in Bangalore district of Karnataka. Problems and issues related to statistics of
horticultural crops were discussed in detail in the symposium organized during ISAS Conference
in 2001. Details of the sampling methodology for estimation of extent of cultivation and
production of fruit and vegetable crops are given below:
I.60
the selected orchards, three clusters of four trees each of bearing age may be selected at random
for recording data on yield of a fruit throughout the harvesting season.
2.1.1.2 Sample Size
A total of 150-200 reporting villages (psu’s) may be selected in the district. As described above,
this number may be allocated to different strata (tehsils) in proportion to area under orchards and
the allocated number of villages in a stratum may be selected with pps with replacement. At the
second stage of sampling, 5 orchards may be selected at random and from each selected orchard,
three clusters of 4 bearing trees may be selected at the ultimate stage of sampling. Earlier surveys
have shown that with this type of design and sample size, the average yield at the district level is
likely to be estimated with a Standard Error(S.E.) of about 5% and the area and total production
with a S.E. between 5 to 10%. However, the efficiencies of various estimators would depend
upon the amount of variability in different characters. Surveys conducted during initial years will
provide an idea about these variability and accordingly the number of villages and orchards
selected may be modified to achieve the desired degree of precision.
2.1.2 Sampling Plan for Estimation of Extent of Cultivation and Production of more than
one Fruit Crops in a State
The important fruit crops whose production is to be estimated should be identified first.
Normally, the previous years’ area figures under different fruit crops are available at the
tehsil/taluk level and these may be used to determine the important fruits in the State. Since the
cultivation of fruits is usually not so evenly spread and may in fact be concentrated in a few
districts/regions, the first step in the planning of fruit survey is to identify and delimit the
important fruit growing regions or areas for different fruits. A district is considered too large a
unit of area for this purpose. However, taluks or sub-divisions or equivalent areas in a district
may be considered appropriate. Thus, taluks which are important at least for one of the fruit
crops, may be identified as important fruit growing taluks. It may be mentioned that importance
of a taluk with respect to a fruit is determined on the basis of area under that fruit and thus a
taluk important for a given fruit may not be important for other fruits. As a broad guideline, for a
given fruit, the important taluks are those which taken together cover 40-50% of the total area
under that fruit in the entire State.
2.1.2.1 Sampling Design and Sample Size
All taluks/sub-divisions, considered important fruit growing areas as described above, may be
taken as strata. The remaining area or taluks may be further classified or grouped into 4 to 5
strata with respect to importance of individual fruit crops taking into account the geographical
contiguity. In these strata, taluks may be considered as primary sampling units. Thus survey
would then cover all important fruit growing taluks i.e.taluks in which fruit cultivation is
concentrated as well as the selected taluks out of the rest.
In the selected taluks also, all the villages may not be growing all the fruits. A frame of villages
growing different fruit in a stratum is, therefore, prepared. Accordingly, villages in a stratum
may be classified into two categories (i) growing at least one fruit and (ii) growing no fruit at all.
In category (i) on the basis of village-wise area under fruits, villages may be identified as
‘reporting’ or “non-reporting” for individual fruits. If the reported areas are considered as
reliable, efforts may be concentrated only in the reporting villages for each fruit. However,
experience shows that faulty reporting is not uncommon and therefore, adequate representation
I.61
may be given to non-reporting group. From the reporting group of villages for a given fruit crop,
four villages may be selected with replacement and with probability proportional to area reported
under the fruit crop. From the non-reporting group of villages (in which other fruits are grown), a
sample of two villages may be selected in each stratum by SRSWOR. From the villages in
category (ii) where no cultivation of fruits is reported, a sample of two villages may be selected
by SRSWOR. The selected villages may be completely enumerated for the extent of cultivation
and number of trees in orchards and also the stray trees.
For yield estimation, a sub sample of two villages out of four reporting villages may be retained
in all the major fruit growing taluks/strata and from each village 5 orchards and 3 clusters of 4
trees each of bearing age may be selected for this purpose. The selected clusters of trees may be
observed for entire harvest period both with respect to weight as well as number of fruit.
However, exceptions to this procedure may be made for certain crops like banana and grapes. A
uniform approach in this regard is essential for comparability as well as pooling of estimates
over different areas.
2.2 Vegetable Crops
The survey approach for estimation of area and production of vegetable crops is somewhat more
complex due to special feature of cultivation of these crops. Some of these features are as
follows:
i. The vegetables are short duration crops and their duration varies considerably from one
vegetable to the other.
ii. Harvesting of vegetables involves a number of pickings
iii. Vegetable cultivation is more or less a continuous process with various operations like
sowing, harvesting, etc. being done simultaneously in different fields of a village.
iv. Vegetables are highly sensitive crops and this normally adds to the variability in the yield
rates of the crops.
It is also realized that due to perishable nature of the vegetable crops, production depends on
availability of marketing facilities in the area. This is why cultivation of vegetables is normally
concentrated around bigger town and cities. Accordingly, the methodology for estimation of area
and production of vegetable crops was developed at the district level in different surveys
conducted in various states.
The sampling design for surveys for estimation of area and production of vegetables is described
below:
2.2.1 Sampling Design
The sampling design is a stratified multistage random sampling. Taluks or equivalent areas may
be taken as main strata. Further, since area under vegetables may vary considerably from one
village to another in a taluk, sub-stratification may be done on the basis of village-wise area
under vegetables. For this purpose, 3 to 4 substrata with equal area under vegetables may be
formed. The data figures may be available in revenue records. If not available, then a preliminary
survey may be conducted to obtain village wise area under vegetables. Within the strata, clusters
of three villages may be taken as primary sampling units. For determining the extent of
cultivation, a sampling fraction of about 20% may be used for selection of clusters of villages.
I.62
The allocation of clusters of villages to different strata may be done in proportion to area under
vegetables. The allocated number of clusters in different strata may be selected with SRSWOR.
For yield study, 50% of the clusters selected for area may be retained and fields growing
vegetables may be selected in these clusters. The selected clusters of villages may be completely
enumerated for area under vegetables. Vegetables being short duration crops, one time
enumeration in a year may not be meaningful. To account for the short duration of crops and
early and late varieties, a year may be divided into four periods of three months each. The area
enumeration may be done in the beginning of each period. This will also provide a frame of
vegetable fields for estimation of yield rate. For estimation of production, 6 to 8 fields of each
important vegetables may be selected in each of the clusters selected for yield study. In each of
the selected fields, a randomly located crop cut plot of 5m x 5m may be demarcated and
observed for all the pickings in the respective periods. The yield of a vegetable for a selected
field is obtained as the aggregate of all pickings in the period obtained from the c.c. plot. The
average yield of the vegetable for the village is obtained as a simple mean of field wise yield and
when multiplied by the area under vegetable in the village gives the vegetable production in the
village. In this way the production for each period may be estimated separately. The average
yield is then obtained from the estimated production and the area under a vegetable.
This sampling design is likely to provide estimates of average yield with less than 5% standard
error and the area and production with less than 10% standard error for important vegetable
crops at the district level.
3. Present Status
Country level estimates are developed only for important fruit and vegetable crops under central
scheme "Crop Estimation Survey on Fruits and Vegetables". Although this scheme is in
operation for the last several years, coverage in terms of fruits and vegetables as well as in terms
of geographical coverage of the country, it is grossly inadequate. Since DES releases the
estimates for only eleven State, users are depending on other sources of statistics. National
Horticulture Board (NHB) of Ministry of Agriculture is bringing out the publication entitled
"Indian Horticulture Database”. This publication contains recent data on area, production and
prices of various horticultural produce. NHB data is relatively more comprehensive by way of
covering data for most of the states, and is generally used for the purpose of GDP estimates etc.
The estimates released by NHB are not based on sound statistical methodology and need to be
examined.
It is evident from the above that reliable and timely estimates of area and production at all India
level for these crops are not available from any source. Moreover, there are problems in the
implementation of the pilot scheme on fruits and vegetables because of various reasons.
4. Recommendations of National Statistical Commission (NSC)
National Statistical Commission (NSC) - 2001 in its recommendation (refer 4.5.7 on page 122)
has suggested that:
i. “The methodology adopted in the pilot scheme of Crop Estimation Survey on Fruits and
Vegetables should be reviewed and an alternative methodology for estimating the
production of horticultural crops should be developed taking into account information
flowing from all sources including market arrivals, exports and growers associations.
I.63
ii. Special studies required to establish the feasibility of such a methodology should be taken
up by a team comprising representatives from IASRI, DES, Ministry of Agriculture
(MoA), Field Operations Division (FOD) of National Sample Survey Organization
(NSSO) and from one or two major states growing horticultural crops.
iii. The alternative methodology should be tried out on a pilot basis before actually
implementing it on a large scale”.
5. Alternative Methodology
In view of NSC recommendation, a Brainstorming Session on "Alternative Methodology for
Estimation of Production of Horticultural Crops" was held on 24th November 2003 at IASRI,
New Delhi under the Chairmanship of Economic and Statistical Advisor to discuss the issues
regarding strengthening of horticulture database.
Accordingly, a project proposal entitled “Pilot Study to Develop an Alternative Methodology for
Estimation of Area and Production of Horticultural Crops" after incorporating the suggestions
made during the brainstorming session was submitted by IASRI to DES, MoA for funding. The
project was finally funded by Central Statistical Organization (CSO) Ministry of Statistics and
Programme Implementation and was completed at IASRI. This study was conducted in two
States namely, Maharastra and Himachal Pradesh covering important fruits and vegetables.
5.1 Sampling Design and Sample Size
First of all, important districts were identified for conducting survey on the basis of district-wise
area figures under fruits and vegetables of the State. As a broad guideline, the important districts
are those which taken together cover 70-80% of the total area under fruits and vegetables in the
entire State. The sampling design which was adopted for the survey may be described as
stratified multistage random sampling. Taluk-wise area figures under fruits and vegetables were
used for stratifying the taluks of the selected districts into two groups viz. high productive taluks
and low productive taluks. High productive taluks are those which constitute 60-70 % of the total
area under fruits and vegetables of the district and rest of the taluks will fall under low
productive taluks. A sample of two taluks was selected by SRSWOR from both the groups after
rejecting taluks contributing less than 5% of total area under fruits and vegetables of the district.
From each of the four selected taluks, a sample of five villages was selected by SRSWOR. The
selected villages were completely enumerated so as to record number of orchards under different
fruits and cropping pattern with respect to vegetables. An orchard for selection process should
have minimum of 12 fruit trees of bearing age of a single fruit crop. For fruits survey, a sample
of five orchards was selected from each selected village by SRSWOR. In case, there are more
than one fruit crop available in the village then orchards were selected in proportion to the
number of orchards for two major fruit crops in each of the village with a minimum of two
orchards for each fruit crop. Major fruit crops were decided on the basis of number of orchards
of different fruits available in the village. From each selected orchard, a sample of three clusters
each consisting of four trees of bearing age was selected randomly out of the total number of
trees of bearing age. The yield of selected trees was collected through enquiry and yield of any
four trees was collected through physical observation.
For vegetable survey, a sample of 10 vegetable growers was selected out of qualified vegetable
growers of a village. For this, after complete enumeration of selected village, a list of qualified
vegetable growers was prepared. Qualified growers are those vegetable growers who have 0.1 ha
and above gross cropped area under vegetables in case of Maharashtra and 0.01 ha and above in
I.64
case of Himachal Pradesh. Ranking of qualified vegetable growers was done as per grass
cropped area and then qualified vegetable growers were divided into two groups after ranking. If
number of growers is odd, the first group will have one more grower than the second group. A
total of six vegetable growers were selected from the first group and rest four from the second
group. In case total number of qualified vegetable growers in any village is less than or equal to
ten, all the growers were selected for detailed survey enquiry. The produce of all the vegetables
crops grown by the selected vegetable grower was recorded through enquiry and physical
observation was taken on the day of visit. The Field Investigator must get in touch with the
grower of the selected field from time to time and ascertain the date of harvest. He must be
present on the day of harvest. He must locate the experimental plot of specified size (5mx5m)
before the cultivator starts harvesting the field. In each selected field, the experimental plot of the
specified size must be located at random beginning with South-West corner of the selected field.
6. Project Implementation
The major steps followed for the successful completion of the study are as under:
i. Data collection work was done in two phases i.e. complete enumeration for three months
and detailed survey for one year with the help of Commissionerate of Agriculture, Pune
and Directorate of Land Records, Shimla. Seventeen (17) FIs were hired in Maharashtra
and eight(8) were hired in H.P. with the help of respective State Govts.
ii. Schedules along with instruction manual were prepared for data collection. Training for
data collection was provided to the hired FIs in both the states in two phases.
iii. Supervision of data collection was done by the IASRI officials at regular intervals and all
necessary support for this purpose was provided by the State Govts.
iv. The data was coded and entered. The entered data was scrutinized.
v. Estimation procedures were developed for area, production and productivity of fruits as
well as vegetables.
vi. Data analysis was done and estimates of area, production and productivity of important
fruits and vegetables were obtained for surveyed districts of both the states.
vii. The district-wise market arrival data for Maharashtra and H.P. for last ten years for
important fruits and vegetables under the study was obtained from Maharashtra
Agricultural Marketing Board, Pune and from H.P. Agricultural Marketing Board, Shimla
respectively.
viii. The district-wise data for last ten years pertaining to area, production and productivity of
important fruits and vegetables in Maharashtra and H.P. was also obtained from the
respective State departments.
ix. Regression Analysis was done for each district of the State for each major fruit and
vegetable crop using total production and market arrival data for the last 10 years.
x. The total production of important fruits and vegetables for non surveyed districts of both
the states for the year 2006-2007 was predicted using the market arrival data for the year
2006-2007.
xi. State level estimates were obtained by pooling the estimates of surveyed and non
surveyed districts.
I.65
7. Salient Findings
The salient achievements of the study are as under:
i. It is worth mentioning that based on the present study the recommended sample size i.e.
number of villages to be selected from a district is 80. Hence, there is a decrease in
sample size i.e. from 150-200 villages per district to 80 villages per district.
ii. The survey procedures under the alternative methodology have been simplified. These
are cost effective and less time consuming.
iii. The alternative methodology provides estimates for more than one fruit/vegetable at
district level, whereas, the methodology used under CES-F&V provides estimates for a
single fruit/vegetable at district level.
iv. The developed methodology is simple and easy to implement both for fruits and
vegetables.
v. The estimates of number of stray trees have also been obtained and used in finding out
estimates of total production of fruit crops.
vi. The correlation coefficient obtained between production figure and market arrival data is
high.
vii. Market arrival data has been used for obtaining State level estimates.
viii. The percentage standard errors of the estimates obtained for fruits are between 5 to 20
and for vegetables are between 5 to 30 percent at district level based on sample size of 20
i.e. 20 villages per district as per the proposed sampling design.
ix. It is expected that at district level, the total production of important fruits can be
estimated with less than 10 percent standard error and the total production of important
vegetables can be estimated with less than 15 percent standard error at 95 percent
confidence interval, if 80 villages are selected from each selected district.
x. The estimated sample size for obtaining less than 15 percent standard error of the
estimates of total production in case of fruits and less than 20 percent standard error in
case of vegetables at 95 percent confidence interval is 43 i.e. 43 villages per district.
8. Conclusion
This study is the first pilot study in the direction of developing alternative methodology. The
study has revealed very encouraging results as shown above and demonstrated the feasibility of
estimating the production of fruits and vegetables with much smaller sample size. In view of
above, it is recommended that the methodology needs to be tested in few states before it is
implemented at large scale. It may be indicated here that the National Statistical Commission has
also recommended that the alternative methodology should be tried out on a pilot basis before
actually implementing it on a large scale.
Once the alternative methodology is tested, the developed methodology will bridge the data gaps
in official statistics related to horticultural crops. Once the methodology is developed, the greater
commitment and willingness of all concerned to strictly comply with prescribed procedures and
time schedules will bring about remarkable improvements in the system.
I.66
References
1. Cochran, W.G. (1977). Sampling Techniques. 3rd Ed., Wiley Eastern Limited, New
Delhi.
2. Proceedings of the Symposium on “Statistics of Horticultural Crops. Problems and
Issues. J. Ind. Soc. Agril. Statist., 54(1), 119-138.
3. Report of the National Statistical Commission(2001). Ministry of Statistics and
Programme Implementation publication.
4. Singh, D., Manwani, A.H. and Srivastava, A.K. (1976). Survey on fresh fruits in Tamil
Nadu. IASRI publication.
5. Singh, D., Srivastava, A.K., Singh, P. and Pal, S. (1976). Survey on vegetables in
Banglore district of Karnataka State. IASRI publication.
6. Sukhatme, B.V., Manwani, A.H. and S.R. Bapat (1969). Survey on vegetables in rural
areas of Delhi. IASRI publication.
I.67
LIVESTOCK SURVEYS
Hukum Chandra
I.A.S.R.I., New Delhi -110012
1. Introduction
Livestock form an important constituent of the economy of our country in general and that of
agricultural sector in particular. Numerically, the livestock wealth of the country is highly
impressive. India's livestock population according to 2003 livestock census is 187 million
cattle, 96 million buffaloe, 441 million poultry, 62 million sheep, 120 million goat and about
14 million pig besides other livestock like horse, ponie, mule, donkey, camel, etc. A
comparative picture of India’s position in world livestock population, according to 2003
(FAO Production Year Book Vol. 57) data is that India ranks first in case of cattle and
buffaloes, second in goat, third in sheep, fourth in camel and fifth in poultry*. As far as the
production is concerned, India ranks first in milk and fourth in egg production*. Though
India ranks first in the milk production, yet the per capita availability is much below the
nutritional standard. With such a large livestock population, it is expected to have a large
contribution to national income from livestock and livestock products. But because of poor
genetic potentiality of our livestock, the national income derived from their products is about
5 to 6 percent.
The economic importance of livestock and need for increasing production was realised as
early as in the first decade of the last century. Accordingly, steps were taken to improve the
conditions of livestock by organising veterinary and animal husbandry departments in the
States. Thereafter, research in developing good breeds of livestock for improving their
production potential was also initiated. A reliable base for livestock statistics plays an
important role in formulation of various livestock developmental programmes in the country.
Some of the important components of livestock statistics are livestock population and their
products, breed, age composition, feeding and management practices, extent of availability of
improved breeding facilities and veterinary aid availed of, etc. Collection of such statistics
on regular basis help in assessing and evaluating the success and impact of various
developmental programmes implemented in the livestock sector, constraints in their
implementation, if any and their continuous monitoring.
2. Sources of Livestock Data
The only source of livestock statistics in the country prior to 1950 was quinquennial livestock
census which was started in 1919.The last one was conducted in 2003. Presently, the census
provides statistics on age-wise, sex-wise, breed-wise (crossbred and non-descript) number of
animals at one point of time. Since these censuses are normally conducted after every five
years, Intercensal estimates are not available from this source. Sample surveys provide an
answer to such problem. These surveys are also needed to be conducted for collecting data on
additional items not covered in the census.
-----------------------
*Source: FAO Production Year Book - 2003, Volume 57.
Before start of regular surveys, for livestock products, the available official estimates of
production were those obtained by the Directorate of Marketing and Inspection (DMI),
Ministry of Food and Agriculture, Govt. of India through market surveys. These surveys were
not based on objective criteria and as such had limited utility. There was a need to obtain the
estimates based on objectively planned sample surveys. For this purpose, a series of
methodological studies were carried out at IASRI which are described below.
3. Estimation of Livestock Numbers and Products
3.1 Estimation of Livestock Numbers
For obtaining reliable estimates of livestock population, IASRI initiated a number of pilot
investigations for evolving an appropriate sampling technique for estimation of livestock
numbers and also providing a plan of rationalised supervision for the census work to improve
the quality of the data in the districts of Etawah (Uttar Pradesh) and Wardha of Madhya
Pradesh (now in Maharashtra) during 1951 and 1953 respectively. The design evolved under
these two pilot investigations was tested on wider scale in whole of the former Bombay State
during 1954 and results of the survey demonstrated the efficacy of the sampling approach for
estimation of livestock numbers.
3.2 Estimation of Livestock Products
During the Second Five Year Plan (1956-1961), Institute also conducted a series of pilot
surveys for evolving suitable sampling technique for estimation of output of major livestock
products viz. milk, egg, wool and meat on individual basis. The regions covered for
estimation of milk production were Punjab plains, Eastern Uttar Pradesh, Gujarat and coastal
districts and adjoining areas of Andhra Pradesh and also the coastal districts of Orissa. For
estimation of egg production, the States covered were Andhra Pradesh, Kerala and West
Bengal; for estimation of wool production, the States covered were Gujarat, Rajasthan,
Mysore (now Karnataka), Himachal Pradesh and Andhra Pradesh whereas for estimation of
meat production, the States of Tamil Nadu and Haryana were covered. In the Third Five Year
Plan (1961-1966), some tracts covered during Second Plan Period were repeated to test the
soundness of the techniques developed and also to study the changes in the level of output
and total production of these principal livestock products. These surveys clearly demonstrated
the feasibility of estimating the individual products, through well planned sample surveys.
However, it required a lot of infrastructure to conduct these surveys regularly, if continuity of
data was to be maintained. Such a system could provide yearly estimates with seasonal break
up also. But it was felt that if the estimates of individual products are to be estimated
regularly and the emphasis is also to be laid on estimation of changes taking place over time
(such as yearly changes or changes over plan periods) then the surveys may be planned in an
integrated way. Pilot sample surveys with this concept of integrated approach were carried
out in Northern region comprising the States of Punjab, Haryana and Himachal Pradesh
during 1969 -72 and Andhra Pradesh of southern region during 1971-74 for developing a
sampling technique for simultaneous estimation of all the principal livestock products in one
single survey.
An essential feature of these surveys was that the stratification of each region was done on
the basis of taluks important for a particular product viz. important for poultry alone,
important for sheep alone and important for both poultry and sheep. In southern region taluks
important for both poultry and sheep were further divided into two sub-strata viz. important
for poultry and mutton type sheep & poultry and wooly type sheep. In each year, one product
was covered on intensive scale whereas the other products on smaller scale. Milk, poultry and
egg production, wool and meat production were covered on larger scale in first, second and
I.69
third year respectively. In three seasons of each year, successive sampling procedure was
strictly followed whereas over the years this pattern was not adopted as no independent
unmatched samples were selected except in first year. Thus, in order to obtain an improved
estimate of mean of the character (say milk) per season in the first year successive sampling
methodology was used and obtained similar estimates for subsequent years and double
sampling procedure was utilised by using information collected on large scale in the first
year. The studies undertaken in these two regions demonstrated the feasibility of obtaining
reliable estimates of these four products simultaneously in a single survey if we strictly use
the suggested design.
This approach, however, had a limitation that the estimation was rather complex and in the
actual follow up, the approach was simplified to the extent that it could be implemented
conveniently through the State Departments.
4. Sampling Design Adopted and Sample Size Covered Under Integrated Sample
Surveys
The sampling design being adopted for the surveys for the estimation of production of milk,
egg, wool and meat is a stratified multi-stage random sampling with villages as the first stage
unit, households/cluster of households as the second stage unit and the animals within the
households as the third and ultimate unit whereas no sub-sampling of layers within a
household is done for recording data on egg production. The animal husbandry districts in the
states are taken as strata. For the estimation of livestock numbers a sample of 15% of the
villages are selected in the State for complete enumeration of livestock population (5%
villages in each season viz. rainy, winter and summer). The sample of villages in each season
is allocated to different strata in proportion to the population of livestock in them. From the
selected villages, a representative sample of 10-12 villages is selected for collection of
detailed information for the estimation of district level estimates of milk, egg, wool and meat.
The sample of 10-12 villages is allocated to different tehsils/group of tehsils which
constitutes a sub-stratum according to livestock population in them. The selection of second
stage units was done with equal probability and without replacement with sample size as
follows:
1st Round
Milk: 2 clusters of 2 households each
Egg: 2 clusters of 5 households each
Wool: Sample of 5 households (flocks)
Meat: 2 recognised slaughter houses
2nd , 3rd and 4th Rounds
Milk: 4 clusters of 2 households each
Egg: 4 clusters of 5 households each
Wool: Sample of 8 households (flocks)
Meat: 2 recognised slaughter houses
The recording of wool yield will be done in the shearing season in the selected villages from
the sample of 5/8 households having sheep. The selection of ultimate unit of sampling was
also done with equal probability and without replacement and the sample size covered was as
follows:
I.70
Milk: Two animals in milk (one cow and one buffalo or both cow or both buffaloe as the
case may be) and all goat in milk.
Egg: All the laying birds (this will include all laying ducks where the same are available)
Wool: Two ram/two wether, two ewe, two lamb
Meat: Three sheep, three goats, three pigs.
For estimation of meat production, an additional sample of two registered slaughter houses
are selected at random in each stratum in a round and the information on meat production are
collected from the sample of three animals of each species viz. sheep, goat, pig and buffaloe.
Type of Data Collected
The type of data from the selected households for all the four major livestock products are
collected on the following items:
(i) Details of livestock and poultry maintained
(ii) Yield rates of livestock products and attendant animal husbandry practices
(iii) Consumption of livestock feeds and their costs
(iv) Number of deaths and causes of death
(v) Protection and treatment against diseases
(vi) Disposal of livestock products and by-products
5. Estimation of Cost of Production of Milk, Poultry and Egg
5.1 Estimation of Cost of Production of Milk
Besides conducting surveys on estimation of production of major livestock products viz.
milk, egg, wool and meat, the Institute also conducted surveys on estimation of cost of
production of milk and egg. The first survey on estimation of cost of milk production was
conducted in and around Delhi during 1953-55 and in and around Madras and Calcutta cities
during 1957-59 and 1960-62 respectively to test soundness of the methodology developed in
Delhi survey. In these surveys data were recorded at weekly intervals and covered both
commercial and private producer households. The deeper study revealed that it will be more
appropriate if only commercial producer households are kept under observation for
estimating the production cost and data be collected at fortnight interval. These observations
were taken into consideration while developing the methodology for simultaneous estimation
of availability of milk and its cost in milk shed areas of milk supply schemes. For this
purpose a series of surveys were conducted in Krishna Delta area of Andhra Pradesh (1975),
Dhulia region of Maharashtra (1978) and I.C.D. area of Bhopal, Madhya Pradesh (1980).
Surveys were planned in such a way that the area under the survey was divided into eight
sectors on the basis of number of milch animals and geographical contiguity. In each sector
four clusters of three villages each (p.s.u.'s) were selected and out of four p.s.u.'s, two p.s.u.'s
were retained throughout the period of enquiry for cost study whereas the other two were for
the study on availability of milk and selected afresh during each season. The cost of milk
production was obtained by pooling the cost involved in feeding and management of the
animal, labour (hired as well as family labour), depreciation on animal and on capital
investment, interest on capital, etc. minus the income from dung produced by the animal.
5.2 Estimation of Cost of Production of Poultry and Egg
The Institute conducted surveys for estimating of cost of poultry and egg production
separately under commercial management i.e. farms having at least 50 layers and habitually
I.71
sold egg and/or bird as well as under small scale farming i.e. the farmers maintaining 10 to 15
birds in the households. Under commercial management, two pilot surveys were undertaken
in Tanda and Dasuya areas of Hoshiarpur district of Punjab during 1967-69 and in around
Delhi during 1969-71. Similarly, under small scale farming, four taluks viz. Warangal,
Mahbubnagar, Janagaon and Narsampet of Warangal district in Andhra Pradesh during 1979-
80 were covered.
In cost of production of poultry and egg study, a large number of villages were selected for
preliminary enumeration and for detailed enquiry a sub-sample of villages having poultry
farms was selected. In addition to these villages, urbanised localities having at least one
commercial poultry farm were also included in the study. For obtaining cost of egg
production, certain cost components viz. feed, labour, depreciation of bird, assets and
equipments, interest on capital cost of maintaining an adult bird and other expenses were
included in the study. In order to have the overall cost, the information on miscellaneous
income from disposal of gunny bags, poultry manure, etc. was also collected. The cost of
production per egg was calculated from selected villages and commercial poultry farms
separately.
It is evident that there is basic difference in the methodological approaches for estimation of
production of major livestock products and cost of production of these products. In estimation
of production of livestock products same p.s.u.'s (villages) are observed and s.s.u's
(households) vary from round to round in a particular season and information on number and
yields are collected whereas in cost of production surveys same p.s.u.'s and s.s.u.'s are
observed throughout the period of study and information on cost aspects are collected. Since
the sampling units and information to be collected under these studies are entirely different so
by clubbing these two in a single survey, the enumerator finds it difficult to collect entire
information on these aspects and he is likely to loose some information on either of the study.
Besides this, there are other problems like non-cooperation from the villagers in supplying
data needed by actual weighment by the enumerator, etc.
Keeping these facts in mind, it is advisable not to club these two studies in a single survey
and separate staff may be provided for each study so that reliable estimates can be prepared.
6. Role of Technical Committee of Directions (TCD)
On the basis of these pilot investigations, a workable methodology was available for conduct
of surveys for estimation of livestock products and as well as cost of production of important
livestock products. During the Fifth Plan Period, Government of India initiated a centrally
sponsored scheme for establishing and strengthening statistical cell in Animal Husbandry
Directorates of various states and union territories for conducting sample surveys for
estimation of livestock products. A Technical Committee of Direction (TCD) for
improvement of Animal Husbandry and Dairying Statistics was also set up in which the
representatives from the Centre and the States/UTs were appointed as the members of the
committee during 1976. This was an important step towards smooth implementation and
conduct of these surveys. It provided a platform for continuous monitoring and improvements
in the data collection system of livestock statistics. A sub-committee was also constituted by
this committee to review the methodologies developed by IASRI for estimation of these
products and the one being followed in some of the states.
The sub-committee noted that all India estimates of production of these products could not be
built up because the regular surveys on any one product were not being carried out in all the
states at one point of time. The other shortcomings were lack of uniformity in coverage,
questionnaires being used, proper supervision, tabulation and analysis of data. It was
I.72
suggested that the surveys should be conducted every year in all the states and union
territories for all the products simultaneously on uniform pattern. Keeping in view the
simplicity and field convenience, necessary modifications were also suggested to be made by
IASRI so that these can be carried out on regular basis and work be coordinated by statistical
cell created in the Ministry of Agriculture at the centre. The Institute made an effort to
simplify the sampling plan, estimation procedure for estimation of different livestock
products and also for estimation of cost of production of milk and egg and schedules to be
canvassed under this scheme and passed on to the Department of Animal Husbandry and
Dairying, Ministry of Agriculture for smooth implementation of this scheme by the State
Animal Husbandry Departments of various states on regular basis.
The main task of the TCD constituted during 1976 is to streamline the statistical activities,
identify data gaps and to provide technical guidance in the statistical programmes relating to
animal husbandry and dairying. This committee meets on regular intervals with the
representatives of State Animal Husbandry Departments under the Chairmanship of Director,
IASRI and seven such meetings have been organized to approve the state-wise estimates of
major livestock products viz. milk, egg, wool and meat prepared by the State Animal
Husbandry Departments on the basis of the surveys conducted by them and thus to build up
all India Estimates. The committee also discusses the difficulties being faced in data
collection and suggests the remedies. It also identifies the data gaps and suggests the suitable
measures to be taken up by state departments and other organizations engaged in this field.
The committee also discusses the issues like conduct of livestock census, uniformity in data
collection and the reporting of annual estimates on uniform basis in time. During the previous
TCD meeting some problems and suggestions have been put forward. One of the important
points has been that the estimates of milk and egg are obtained for all the states but for meat
and wool there have been some problems.
As regards for building up wool production estimates some states show their inability to do
so because of non-existence of sheep in the selected villages of certain tracts and also the
areas which have sheep population, the enumerators fail to record the wool yield data in their
presence even after regular follow up. In order to overcome these problems, TCD suggested
to identify some pockets in the state having the sheep population before the start of the survey
and efforts should be concentrated in those areas only. Secondly, the data on wool yield may
also be collected from various shearing centres, sheep breeding farms in addition to the
selected flocks in the villages.
Regarding meat production, major portion of meat is produced in the slaughter houses and
since slaughter houses come under the preview of local self-government and there is no check
by the officials of the animal husbandry departments, so it is difficult to collect information
on meat production and also the reliability of the number of animals slaughtered there as
given in their records is not ensured. Secondly, there is no clear-cut concept and definition of
slaughter houses and varies from state to state. TCD recommended having a fresh list of
slaughter houses on the basis of some uniform concept and definition for each state. The
committee also recommended to have a sample checking of the data provided by the
slaughter houses, make provision for weighing equipment for weighing the animals in the
slaughter houses and also to collect information on meat production not only from slaughter
houses but also from other sources viz. unregistered slaughter houses (butcher houses) and
sample of households in the selected villages.
The TCD also emphasised to have a quality check of the data being collected by the staff
engaged for this purpose. For this an intensive supervision of work may be done by the
officers of the animal husbandry departments and also by the team from the centre and the
I.73
scientists from the IASRI on regular basis. It was also pointed out that the methodologies on
estimation of these products were developed two decades before and since then a lot of
changes have occurred in livestock sector. With the implementation of cross breeding
programme in the country, certain new breeds have come up which have boosted the yield of
these products. Under this changing scenario, it was felt that the reappraisal of the
methodologies being adopted by the states at present might be done by incorporating breed
and other relevant information in the study. This will help in giving correct and reliable
picture of the estimates of production of these products in the country.
Production Estimates of Milk, Egg, Wool and Meat
(2005-06)
States/UTs Milk(000Tns) Egg(Lakh No.) Wool(000Kg) Meat(000Tns)
Andhra Pradesh 7624 164534 3978# 457
Arunachal Pradesh# 48 73 14 20
Assam 747 5359 - 27
Bihar 5060@ 10012 220 176
Chhattisgarh 839 8875 244 4
Goa 56 146 - -
Gujarat 6960 5775 3123 18
Haryana 5299 15125 1136 73
Himachal Pradesh 869 753 1603# 3
Jammu & Kashmir# 1400 6320 7400 27
Jharkhand 1335 6970 150 43
Karnataka 4022 18348 5598 100
Kerala 2063 11956 - 56
Madhya Pradesh 6283 9414 431 19
Maharashtra 6769 35227 1640 236
Manipur 77 835 - 23
Meghalaya 73 973 - 37
Mizoram 15 326 - 9
Nagaland 74 868 - 63
Orissa 1342 12787 - 52
Punjab 8909 35200 712 4
Rajasthan 8713 7029 15405 -
Sikkim 48 146 2 -
Tamil Nadu 5474 62225 750# 119**
Tripura 87 1100 - 12*
Uttar Pradesh 17356 9228 1459 198
Uttarakhand 1206 1873 353 6
West Bengal 3891 29637 666# 487
Andaman & Nicobar 20 477 - 0
Chandigarh 46 280 - 1
Dadar & Nagar Haveli - 48 - -
Daman & Diu 1 11 - -
Delhi 310 196 - 31
Lakshadweep 2# 124# - -
Pondicherry 43 105 - 7
All India 97061 462307 44884 2310
I.74
** unregistered sector also included
- N.A # Projected
@ The milk production of Bihar is considered abnormal due to phenomenal growth of about 70%. The estimate should be again examined
and reasons for the same to be provided.
Source : State/UT Animal Husbandry Departments
I.75
9. Other Related Studies Undertaken by IASRI
1. Estimation of livestock by-products (Hides and Skin)
i) Ludhiana, Amritsar, Jalandhar and Ferozepur Districts of Punjab during 1974-76
ii) Agra and Kanpur Districts of Uttar Pradesh during 1980-81
iii) Chingleput and North Arcot Districts of Tamil Nadu during 1985-86
iv) Surat District of Gujarat during 1985-86
2. Feed intake by bovines through stall feeding and grazing
i) Jhansi District of Uttar Pradesh during 1969-70
ii) Puri District of Orissa during 1978-79
3. Post production losses of milk in rural areas: Estimation of post production losses of
milk in rural area in the district of Rohtak (Haryana)
4. Estimation of production of poultry meat in the district of Gurgaon (Haryana)
Collaborative Studies Undertaken by IASRI
i. Estimation of wool production – emerging data needs and a methodological reappraisal
in the districts of Bikaner of Rajasthan State and Kolar of Karnataka State (AP Cess
Fund)
ii. Assessment of harvest and post-harvest losses (NATP Project)
References
1. FAO Production Year Book - 2003, Volume 57
2. Khatri, R.S., Goyal, J.P., Jayasankar, J. and Geethalakshmi, V. (2005). Estimation of
wool production-emerging data needs and a methodological reappraisal. Project
Report (AP-Cess Fund Project), IASRI, New Delhi-12
3. Khatri, R.S., Goyal, J.P. and Singh, K.B.(1998). Pilot sample survey for estimation of
post production losses of milk in rural areas. Research Report-IASRI, New
Delhi-12
4. Nadkarni, U.G., Somayazulu, L.B.S. & Jain, T.B. (1981). Monograph on estimation
of cost of production of poultry and eggs. Research Series - IASRI, New Delhi-12
5. Singh, D., Garg, J.N., Goel, B.B.P.S., Rajagopalan, M. and Singh, K.B. (1977).
Sampling methodology for estimation of milk production, Northern region, (1969-72).
Research Report - IASRI, New Delhi-12
6. Singh, D., Goel,, B.B.P.S., Garg, J.N. and Singh, K.B. (1979). Sampling methodology
for estimation of milk production, Southern region, (1971-74). Research
Report - IASRI, New Delhi-12
7. Singh, D., Goel, B.B.P.S., Garg, J.N., Singh, K.B. and Rajagopalan, M. (1978).
Sampling methodology for estimation of egg production and study of poultry keeping
practices in Northern region (1969-72) and Southern region (1971-74) Research
Report - IASRI, New Delhi-12
8. Singh, D., Goel, B.B.P.S., Maini, J.S. and Goyal, J.P. (1979). Sampling methodology
for estimation of wool production in Northern region (1969-72) and Southern region
(1971-74) Research Report- IASRI, New Delhi-12
I.76
9. Singh, D., Maini, J.S., Goel, B.B.P.S., and Bassi, G.S. (1978). Sampling methodology
for estimation of meat production in Northern region (1969-72) and Southern region
(1971-74). Research Report - IASRI, New Delhi-12
10. Singh, D., Murty, V.V.R. and Goel, B.B.P.S. (1970). Monograph on estimation of
milk production. Research Series - IASRI, New Delhi-12
11. Singh, D., Goel, B.B.P.S., Garg, J.N. and Rao, D.V.S. (1975). Monograph on Sample
Survey Techniques for Estimation of Egg Production. Research Series - IASRI, New
Delhi-12
12. Singh, D., Rajagopalan, M. and Maini, J.S. (1972). Monograph on estimation of wool
production, Research Series - IASRI, New Delhi-12
13. Singh, D., Rajagopalan, M., Maini, J.S. and Singh, K.B. (1978). Monograph on
sample survey techniques for estimation of meat production. Research Series - IASRI,
New Delhi-12
I.77
BASIC PRINCIPLES OF REMOTE SENSING
B. C. Panda
I.A.R.I., New Delhi - 110012
1. Introduction
Remote Sensing as it has been accepted today in the scientific literature is comparatively a young
branch of science. But the act of remote sensing is perhaps as old as the origin of life on our planet.
Remote sensing is the sensing of an object or a phenomenon from a remote distance. But then, how
remote is remote for remote sensing? Consideration of this type is highly relative and depends on the
characters of the signals and the sensors, and also on the attenuation properties of the signal
transmission channel. One may also like to know whether there can be sensing without physical
contact between the sensor and the object. You may say yes, but the real answer to this question is a
big NO. This is because of the fact that under all circumstances the objects to be sensed and the sensor
are always intimately bathed in an interacting field, namely the gravitational field, the electromagnetic
field and / or the pressure field. These fields are not fictitious but are as real as a lump of matter is.
Thus, whatever may be the distance between the sensor and the sensed, they are always in contact
with each other through the field. What then is the special significance of the contact sensing? In fact,
during the so-called contact sensing no true material contact is established since the planetary /
surface electrons of the two bodies can not touch each other, being both negatively charged they repel
one another and keep themselves at a distance. What happens in reality is that during the so-called
contact sensing the fields of both the bodies influence each other so markedly that it results in an
appreciable amount of interaction force / pressure, which is sensed or measured. Now, without going
any further with the passionate philosophical discussion on what is remote sensing, we may content
ourselves saying that practically remote sensing is the science and technology for acquiring
information about an object or a phenomenon kept at a distance. Basically it is a non destructive
physical technique for the identification and characterization of material objects or phenomena at a
distance.
2. Essential Components of Remote Sensing
Essentially remote sensing has three components:
(i) The Signal (from an object or phenomenon)
(ii) The Sensor (from a platform), and
(iii) The Sensing (acquiring knowledge about the object or the phenomenon after analysis of the
signals, received by the sensor, at the user’s laboratory).
However, the interaction of the signal with the object by which we obtain information about it, and
the interaction of the signal with the transmission channel which reduces the signal strength are given
due considerations for detail information extraction. Remote sensing is a branch of Physics, namely
Reflectance Spectroscopy which has now found extensive applications in almost every field of human
activity.
2.1 Natural and Artificial Remote Sensing
Sensing in general and remote sensing in particular can be taken as a measure of life and activity of all
living organisms, from microbes to man. Any living organism whose sense organs are well developed
can interact with its environment better, live a better life and can protect itself better from its enemies
and hostile environments. Taking the example of man, we have five well developed sense organs
(natural sensors): eye, ear, nose, skin and tongue along with highly developed sensing systems – the
brain and the nervous system and these are called as natural remote sensing.
Seeing is believing. But we human beings can see only through the visible light which forms only a
very narrow band out of the extremely broad electromagnetic spectrum. This is because our eye is not
sensitive to the wavelengths below the violet and above the red region of the electromagnetic
spectrum. Similarly our ear is insensitive to the infrasonic and the ultrasonic frequencies – it can sense
only in the audible frequency range. Thus the knowledge about objects obtained through our eyes and
ears is but partial. With the help of artificial remote sensing, man merely tries to imitate the already
existing remote sensors in us and improve upon them to include wider information channels so that
they can be used efficiently to collect detailed information about objects or phenomena.
2.2 Passive and Active Remote Sensing
A remote sensing system that possesses only a sensor and depends on an external (natural) source to
irradiate the target to be sensed is called a passive remote sensing system. As for example, in visible
light remote sensing system, the sun, an external natural source, irradiates the target and the reflected
light from the target is detected by the sensor. An active remote sensing system, on the other hand,
possesses both the sensor and the source to irradiate the target. As for example, the radar, which is an
active remote sensing system, transmits microwave pulses from its transmitter antenna to irradiate the
target and receives the radar returns by its receiver antenna. Signals are carriers of information. So
before we try to analyze the signals to get the information on the object from which they have come,
we should like to spend some time to study the characters of the signals at some depth. For meeting
the requirements of remote sensing, in general, we can recognize the four types of signals such as
Disturbance in a force field, Acoustic signal, Particulate signal, and Electromagnetic signal. We
would focus on electromagnetic signals only which is major carrier of information in remote sensing
process.
2.3 Electromagnetic Signals
The electromagnetic waves generated by oscillating electric charges all travel with a velocity which is
the highest signal velocity that can be attained. This high signal velocity of electromagnetic radiation
coupled with its low atmospheric attenuation confers a unique positive advantage to the
electromagnetic waves to be used as signals in remote sensing in general and in satellite remote
sensing in particular.
2.3.1 Generation of Electromagnetic Signals
The electromagnetic waves generated by the oscillation of electric charges or fields can have
various frequencies. Thus, the only way these waves distinguish themselves from one another
is through their frequencies of oscillation. If we jiggle the charges or fields with higher and
higher frequencies, we get a whole spectrum of electromagnetic waves generated. For
example, when the frequency of oscillation is of the order of 102 /s as is the case in AC
generators, electromagnetic waves are generated of the frequency 102 Hz which are
recognized as electrical disturbances. Increasing the frequency of oscillation to 105- 106 Hz,
as is achieved in the radio transmitter antenna, we get the electromagnetic waves known as
the radio broadcast band. Similarly by a TV transmitter antenna we get FM-TV waves when
the frequency of oscillation reaches 108 Hz. In electronic tubes such as klystrons and
magnetrons, charges can be oscillated with a frequency of the order of 1010 Hz, and we get
microwaves, radar. For still higher frequencies we go over to the jiggling of atomic and
molecular electrons. In the range of frequency from 1014 to1015 Hz we sense the
electromagnetic waves as visible light. The frequencies below this range are called infrared
and above it, the ultraviolet. Shaking the electrons in the innermost shells of atoms produces
electromagnetic waves of frequency of the order of 1018 Hz, which are known as X-rays.
Increasing the frequency of oscillation to 1021 Hz, the nuclei of atoms emit electromagnetic
waves in the form of -rays. From high energy particle accelerators we can get artificial -
rays of frequency 1024 Hz. At times we also find electromagnetic waves of extremely high
frequency, as high as 1027 Hz, in cosmic rays. Thus we find that the electromagnetic radiation
forms a very broad spectrum varying from very low frequency to very high frequency, or
II.79
from very long wavelength to very short wavelength. Table 1 shows the electromagnetic
spectrum with its broad spectral regions and their rough behavior.
Table 1: Electromagnetic spectrum with its broad spectral regions and their rough
Behavior
Frequency wavelength Name of spectral regions Rough behavior
Hz
100 300,000 Km Electrical disturbance/AC Fields
101 30,000 Km - do - - do -
102 3,000 Km - do - - do –
103 300 Km Audio Waves
104 30 Km - do - - do –
105 3 Km Radio - do –
106 0.3 Km - do - - do –
107 30 m FM – TV - do –
108 3m - do - - do –
109 0.3 m Microwave - do –
1010 3 cm - do - - do –
1011 3 mm - do - - do –
1012 0.3 mm Sub-millimeter / IR - do –
1013 30 m - do - - do –
1014 3 m - do - - do –
0.7 m Visible - do –
0.4 m - do - - do –
1015 0.3 m UV - do –
1016 30 nm X-rays - do –
1017 3 nm - do - - do –
1018 0.3 nm - do - - do –
1019 30 pm - rays (natural) Particles
1020 3 pm - do - - do –
1021 0.3 pm - do - - do –
1022 3 X 10-2 pm - rays (artificial) - do –
1023 3 X 10-3 pm - do - - do –
1024 3 X 10-4 pm - do - - do -
1025 3 X 10-5 pm -rays (cosmic) - do –
1026 3 X 10-6 pm - do - - do –
1027 3 X 10-7 pm - do - - do –
II.80
equilibrium temperature. This is because the radiation which does not escape from the cavity
gets reflected back and forth and fills the cavity. Under this condition the energy lost by the
oscillators in radiation is gained by absorbing energy from the radiation field since they are
illuminated by it. We may say that in the cavity electromagnetic radiation has attained
thermal equilibrium with matter (oscillators). Now if we make a hole in the cavity and look at
it when theoretically its temperature is 0o K, it looks perfectly black. It is for this reason that
the cavity is called a black body, and the cavity radiation the black body radiation. We have
seen that characteristically the black body is a perfect absorber of electromagnetic radiation
and also a perfect emitter. On the contrary, a white body is a non-absorber and non-emitter, it
is a perfect reflector. As a matter of fact, in the natural world we do not come across a perfect
black body or a white body. Natural bodies show a behavior intermediate between a perfect
black body and a perfect white body – the gray body.
2.3.3 Some Characteristics of Black-body Radiation
2.3.3.1 Kirchhoff’s Law
Since no real body is a perfect emitter, its emittance is less than that of a black body. Thus the
emissivity of a real (gray) body is defined by
g = Mg / Mb … (1.1)
where Mg is the emittance of a real (gray) body, and
Mb is the emittance of a black body (perfect radiator).
The emissivity of a black body is b = 1, while the emissivity of a white body w = 0.
Between these two limiting values, the grayness of the real radiators can be assessed, usually
to two decimal places. For a selective radiator varies with wavelength.
2.3.3.2 Planck’s Radiation Law
For a radiating body with emissivity , the spectral radiant (emittance) is given by Planck’s
radiation Law as
M = ( 8hc / 5 ) * 1 / ( ehc/ T - 1 ) , W /(m2 - m) ... (1.2)
= wavelength of radiation
Using the relationship between frequency and wavelength of the electromagnetic radiation
Planck’s radiation law (Eqn. 1.2) can also be written in terms of frequency as
II.81
where = radiation frequency, Hz
A useful feature of Planck’s law is that it enables us to compute the amount of total spectral
radiant emittance which falls between two selected wavelengths or frequencies. This can be
useful in remote sensor design, and also in the interpretation of remote sensing observations.
2.3.3.3 Stefan – Boltzmann Law
If we integrate M or M over all wavelengths or all frequencies, the total radiant emittance M
will be obtained for a radiating body of unit area, i.e.
M = o M d = o M d = ( 854 / h3c3 ) * T4
= T4, W/m2 … (1.5)
where is the Stefan – Boltzmann radiation constant, and has the numerical value
Integrating over the whole spectrum thus provides the total photon flux emitted by the
radiating body of unit area. The Stefan – Boltzmann law for photon flux thus becomes
Q = o Q d = ’ T3 … (1.9)
II.82
where ’ = 1.52 X 1015 m2 s-1T-3. This shows that the rate at which the photons are emitted
by a radiating body varies as the third power of its absolute temperature. For example, for T =
300 K, the total number of emitted photons becomes equal to 4 X 1022 photons / (m2- s).
Thus far we have described the generation of electromagnetic radiation. Among its essential
properties, it does not require a material medium to propagate. In empty space it travels with
the highest signal velocity
c = 2.9979 X 108 m/s
In material medium (non-absorbing or partially absorbing, i.e. transparent or translucent) the
velocity of electromagnetic wave is less, depending on the refractive index () or the
dielectric constant (k) of the medium:
v = c/ … (1.10)
The dielectric constant of the medium is given by
k = 2 … (1.11)
The frequency and wavelength of electromagnetic radiation are related by the formula
= c … (1.12)
The electromagnetic radiation being transverse in character, it can be polarized and its state
of polarization can be changed by reflection, refraction, scattering, absorption and emission.
3. Radiation Terminology
For the measurement of light we commonly use photometric unit which is based on the
assumption that human eye is the ultimate sensor of radiation, and thus the sensitivity of
human eye is the basis for the formulation of these units. In remote sensing, however, sensors
other than the human eye are used to detect radiation in many other parts of the
electromagnetic spectrum to which the human eye is not sensitive. Thus the useful units for
measuring radiation in remote sensing are the radiometric units. Table 2 lists the most
important of these radiometric quantities along with their symbols, defining expressions and
commonly used units.
Table 2: Radiometric and Photometric Quantities
II.83
Irradiance E E = d / dA watt / m2 (Wm-2)
(Spectral irradiance) watt / cm2 (W cm-2)
Radiant exitance / Radiant M M = d / dA watt / m2 (W m-2)
Emittance watt / cm2 (W cm-2)
(Spectral radiant exitance)
Reflectance = r / i - -
Transmittance = t / i - -
Absorptance = a / i - -
Hemispherical reflectance = Mr / E - -
(Spectral hemispherical reflectance)
Hemispherical transmittance = Mt / E - -
(Spectral hemispherical transmittance)
Hemispherical absorptance = Ma / E - -
(Spectral hemispherical absorptance)
3.1 Target– Signal Interaction in Optical Region (Visible and Infrared): Radiation
Modification
An Overall View
The very fact that electromagnetic radiation incident on a target (object) is partly reflected,
partly absorbed and partly transmitted, clearly demonstrates the interaction of
electromagnetic radiation with matter. This fact is summarized as
+ + = 1 … (1.13)
II.84
Now, applying the concept that an ideal absorber is an ideal radiator (emitter), we identify
absorptance with emittance, and write
+ = 1 , or = 1 - … (1.15)
This is Kirchhoff’s law in a different version which shows that for opaque terrestrial material,
the emissivity (absorptivity) and reflectivity (reflectance) are complementary to each other.
Many common terrestrial materials are found to be similar in their thermal emission spectrum
in the 8 – 14 m infrared range with a spectral emissivity between 0.85 – 0.95. Thus for the
measurement of temperature of terrestrial materials by remote sensing technique (with the
help of an IR thermometer), this 8–14m band is invariably used where it is assumed that the
emissivity of all these materials are similar with an estimated emissivity of about 0.9.
3.2 Interaction of Electromagnetic Radiation with Atmospheric Constituents
Electromagnetic radiation from the sun interacts with the atmospheric constituents and gets
absorbed or scattered. Essentially two types of scattering takes place: elastic scattering in
which the energy of radiation ( or ) is not changed due to the scattering, and inelastic
scattering in which the energy of the scattered radiation ( or ) is changed (Compton
scattering and Raman scattering). Inelastic scattering usually takes place when high energy
photons are scattered by free electrons in the ionosphere or by the loosely bound electrons of
atoms and molecules in the atmosphere.
As regards the elastic scattering, we recognize three types of such scattering phenomena in
the atmosphere : Rayleigh scattering, Mie scattering and non-selective scattering depending
on the size of the scatterers in relation to the wavelength of radiation () being scattered. The
atmospheric scattering processes are summarized in Table 3.
= wavelength of radiation
II.85
o = refractive index of the medium
Rayleigh scattering causes the sky to appear blue. Since the scattering coefficient is inversely
proportional to the fourth power of the wavelength, radiation in the shorter blue wavelengths
is scattered much more strongly than radiation in the red wavelengths. The red of the sunset
is also caused by Rayleigh scattering. As the sun approaches the horizon and its rays follow a
longer path through the denser atmosphere, the shorter wavelength radiation is comparatively
strongly scattered out of the line of sight, leaving only the radiation in longer wavelengths,
red and orange, to reach our eyes. Because of Rayleigh scattering, multispectral remote
sensing data from the blue portion of the spectrum is of relatively limited usefulness. In case
of aerial photography, special filters are used to filter out the scattered blue radiation due to
haze present in the atmosphere.
3.2.2 Mie Scattering
Mie scatterers have size . Water vapor and fine dust particles in the atmosphere satisfy
this condition. As a general case, in which there is a continuous particle size distribution, the
Mie scattering coefficient is given by
= 105 a1a2 N(a) K(a,) a2 da … (1.17)
II.86
Table 4: Radiation-matter interaction mechanisms across the electromagnetic spectrum
Spectral Main Interaction Fields of Remote Sensing Application
Region mechanism
-rays Atomic processes Mapping of radio-active materials
X-rays
3.4 Electromagnetic Signals Useful for Satellite Remote Sensing (the Atmospheric
Windows)
The wave bands, narrow or broad, for which the transmission percentage is very high are the
spectral regions that are least attenuated by the atmospheric constituents. These spectral
regions are called the atmospheric windows, and these channels play important role in remote
sensing of the earth resources from satellite platforms. Some of these identified atmospheric
windows are given below in Table 5. From it is seen that the spectral bands of the solar
spectrum which are attenuated due to absorption by the water vapour and other atmospheric
gases, though can not be used in satellite remote sensing of terrestrial surface features yet
they are considered extremely useful for monitoring these absorbing atmospheric gases.
II.87
Table 5: Atmospheric Windows for terrestrial surface monitoring from space-borne
observation platforms
Spectral regions Spectral bands
VIS (Visible) 0.4 - 0.7 m
Spectral bands for remote sensing are chosen from the entire spectrum of the electromagnetic
spectrum so that they can be used for monitoring thematic earth surface features and
phenomena.
4. Sensors and Sensor Platforms
4.1 Sensor Materials
The sensor or detector transforms the energy of the incoming radiation into a form of
recordable information. It is found that no single sensor material is equally sensitive to the
entire range of electromagnetic spectrum. Therefore, different sensor materials are used for
the construction of detectors in different wavelength ranges. In general, there are two types of
electromagnetic signal detectors, namely optical film detectors and opto-electronic detectors.
Sensor materials in the film detectors are silver bromide grains. Usually black and white, true
color and infrared false color films are in use. Black and white and true color films are
sensitive to the visible band of the electromagnetic spectrum (0.4 – 0.7 m). Spectral
sensitivity curves of these photographic films are found to attain a maximum around 0.69 m
and then falls to a very small value at 0.7m. It is because of this that all panchromatic films
operate within 0.4 – 0.7 m spectral band. However, spectral sensitizing procedures can be
applied to the emulsions of black and white films so that they can be made sensitive to near
infrared radiation up to wavelength of approximately 0.9 m. The sensitivity of the false
color film is increased up to 0.9 m including a short but highly reflective part of the near
infrared spectrum (0.7 – 0.9 m).
Opto-electronic detectors are classified into two types on the basis of the physical processes
by which radiant energy is converted to electrical outputs. They are thermal detectors
(sensitive to temperature changes) and quantum detectors (sensitive to changes in the incident
photon flux).
Typical thermal detectors are: thermocouples /thermopiles and thermister bolometers. In
thermocouples and thermopiles, change in voltage takes place due to the change in
temperature between thermoelectric junctions. In thermister bolometers change in resistance
II.88
takes place due to the change of temperature of the sensor materials. Thermister bolometers
usually use carbon or germanium resistors with resistance change of 4 % per degree. Thermal
detectors are slow, have low sensitivity and their response is independent of the wavelength
of the electromagnetic radiation.
4.2 Quantum Detectors
Quantum detectors are of three types. They are:
Photo-emissive detectors (photocells and photomultipliers use alkali metal surface coatings
such as cesium, silver-oxygen-cesium composite)
Photo-conductive detectors (photosensitive semiconductors whose conductivity increases
with incident photon flux), and Photo-voltaic detectors (here modification takes place in the
electrical properties of a semiconductor p-n junction such as backward bias current on being
irradiated with light)
In the wavelength region 1 m, quantum detectors like photo-multipliers and Si-
photodiodes are found quite efficient and do not need cooling devices. However, for satellite
platforms the photo-multipliers are found less suitable because of its accompanying high
voltage power supplies that increase the weight of the payload substantially. In 1 to 3 m
wavelength region, Ge-photodiode, Ge-photo-conductor, InSb photodiode, InAs photodiode
and HgCdTe photodiode can be used for sensor materials. However, these sensors are to be
kept cooled as their electrical characteristics change with increase in their temperature. In the
thermal wavelength region (10 to 14 m), HgCdTe photo-conductors and PbSnTe
photodiodes can be used as sensor materials with effective cooling systems. For the
microwave region of the electromagnetic spectrum, the sensor or detector invariably used is
the microwave antenna.
4.3 Sensor Systems
In a sensor system the sensor material is integrated into its appropriate circuitry and housing
to detect and process the input signals and give out the corresponding outputs for further
analysis to generate information on the target surface from which the signals are received.
Sensor systems are of two types : non-imaging sensor system and imaging sensor system.
Non-imaging sensor system include sounders and altimeters for measurement of high
accuracy locations and topographic profiles, spectrometer and spectroradiometer for
measurement of high spectral resolution along track lines or swath, and radiometers,
scatterometers and polarimeters for high accuracy intensity measurements and polarization
changes measurements along track lines or wide swath.
Imaging sensor systems are again of two types: framing systems and scanning systems. In
framing systems images of the targets are taken frame by frame. These include imagers like
photographic film cameras and return beam videcon. The scanning systems include across
track scanners and along track (push broom) scanners. Imagers and scanning altimeters /
sounders are used for three dimensional topographic mapping. Multispectral scanners /
thematic mappers are used for limited spectral resolution with high spatial resolution
mapping. Imaging spectrometers are meant for high spectral and spatial resolutions. Imaging
radiometers and imaging scatterometers (microwave) are used for high accuracy intensity
measurement with moderate imaging resolution and wide coverage. Brief descriptions of
some of the important framing and scanning systems mentioned above are presented in the
following paragraphs.
II.89
4.4 Framing Systems
4.4.1 Large Format Camera (LFC)
This high performance photographic film camera was used for the first time in Space Shuttle
flight mission of October, 1984. Large format camera has an achromatic lens (in the
wavelength range of 0.4 to 0.9 m) of focal length 305 mm. Its format is 230 X 460 mm and
exposure range is 3 to 24 ms. The camera is flown with the long dimension of its format in
the flight direction to obtain the necessary stereo coverage for the topographic map
compilation. The magazine capacity of the large format camera is 1200 m of film for 2400
frames.
4.4.2 Hasselblad Camera
This was the multi-band camera flown on Apollo-9 flight mission of March,1969 to obtain
the first multi-band pictures of earth from space in the NASA program. It had in the assembly
four Hasselblad cameras with filters in the green, yellow, red and deep red regions of the
electromagnetic spectrum. It was this Hasselblad camera with infrared ektachrom film which
was used to obtain the first historic aerial photograph of the coconut plantation of Kerala in
India in late nineteen sixties. The analysis of the coconut crown intensity showed its strong
relationship with the coconut production by the plants. Investigating the cause of low crown
crown intensity it was found that such plants were affected by a viral disease attributed to the
coconut root wilt virus which was confirmed from electron microscopic studies. The clear
practical demonstration of the fact that the onset of a disease could be identified by the near-
infrared aerial photography, much before its visual symptoms become apparent, gave a strong
boost to the development of remote sensing program of the Indian Space Research
Organization, Department of Space, Government of India.
4.4.3 Return Beam Vidicon (RBV)
Return beam vidicon is a framing system which is an electron imager and works like a
television camera. When the shutter is opened and closed, its optical system frames an image
on a photoconducting plate and retained as a varying charge distribution just like a
photographic a photographic camera frames the image on a photochemical film. The pixels of
the image receiving higher intensity of light becomes more conducting, meaning thereby that
more of the electrons from their corresponding backside pixels move to the front side. Thus
the backside pixels of the photoconducting plate becomes positively charged to different
amounts. Now a fine electron beam from the electron gun is allowed to scan the pixels line by
line. Thus while the electron beam neutralizes the pixels of the photoconductor from the
backside, a part of the beam, now weaker than the forward scanning beam, returns back to a
detector carrying the image describing signals. The strength of these feeble signals is
increased in an electron multiplier and final signals are digitized and recorded as the image
output. These electronically processed image data are amenable to rapid transmission from
sensor platform to a ground receiving station. Being a framing device, the return beam
vidicon collects a full frame data practically instantaneously. Alternately, the images may be
recorded on a magnetic tape for replay at a later time when the sensor platform is within the
range of a ground receiving station.
Landsat-1 and -2 carried three RBVs with spectral filters to record green (0.475 – 0.575 m),
red (0.580 – 0.680 m), and photographic infrared (0.698 – 0.830 m) images of the same
area on the ground. During image processing these data are taken as due to blue, green and
red bands to develop the false color image of the scene. Landsat-3 carried two RBVs
operating in the panchromatic band (0.505 – 0.750 m) to obtain higher resolution (24m)
compared to its MSS ground resolution (79m). Moreover, each of these two RBVs has a
II.90
covered adjacent ground area of 99X99Km with 15Km sidelap. Successive pairs of RBV
images also have a 17Km forward overlap. Thus two pairs of RBV images cover an area of
181X183 Km which is equivalent to the area corresponding to the MSS scene (185X185
Km). The time sequence of RBV scene acquisition was kept at 25 seconds, because of the
driver for the framing sequence of the multi-spectral scanner of the Landsat series of remote
sensing satellites.
4.5 Scanning Systems
4.5.1 Across Track Multispectral Scanner (MSS)
This form of imaging is used in Landsat series of satellites. The scanning system employs a
single detector per band of the multispectral signal. It has an electrical motor, to the axel of
which is attached a solid metal cylinder whose free end is cut at 45 degrees to the axis of its
rotation and highly polished to act as a scanning mirror. The field of view (FOV) is restricted
by an aperture so that the mirror will receive signals in almost nadir view from 2000 ground
resolution cells that makes one scan line. The signal received by the rotating scanning mirror
from a ground resolution cell (corresponding to GIFOV) is a white one and contains spectral
information in different bands. This white beam is reflected by the mirror in the flight
direction (parallel to the ground) and is allowed to pass through a monochromator /
spectroscope which splits the composite beam into its color components. The detectors with
their designed apertures are so placed that they now receive the spectral information from the
ground resolution cells in the specified bandwidths of various color components of the white
signal. The current produced in the sensor material (Si- photodiodes etc.) by the different
bands of the electromagnetic spectrum are digitized and transmitted to the ground receiving
stations or stored in magnetic tapes till a ground receiving station is within its range for
transmission / reception. Since the rotating mirror scans the swath line by line perpendicular
to the track (flight direction), the scanner is called across-track scanner.
The dwell time of the scanner is computed by the formula:
Dwell Time = ( Scan rate per line ) / ( Number of ground resolution cells per line )
The Spatial Resolution of the Scanner = Ground Resolution Cell = GIFOV
4.5.2 Along Track Multispectral Scanner / Push Broom Scanner
In this type of scanner, the scan direction is along the track (direction of flight) and hence the
name along track scanner. It is also called push broom scanner because the detectors are
analogous to the bristles of a push broom sweeping a path on the floor.
Development of charge-coupled device (CCD) has contributed to the successful design of the
along track scanner. In this the sensor elements consist of an array of silicon photodiodes
arranged in a line. There are as many silicon photodiodes as there are ground resolution cells
(corresponding to IFOV) accommodated within the restricted FOV of the sensor optics. Each
silicon photodiode, in turn, is coupled to a tiny charge storage cell in an array of integrated
circuit MOS (metal oxide semiconductor) device forming a charge coupled device (CCD).
When light from a ground resolution cell strikes a photodiode in the array, it generates a
small current proportional to the intensity of light falling on it and the current charges the
storage cell placed behind the diode. The charged cells form a part of an electronic shift
register which can be activated to read out the charge stored in the cells in a sequential
fashion. The output signals are correlated with the shift pulses, and digitized to reconstitute
the image.
Because of the linear array of the sensor elements in this type of imaging / scanning system, it
is also called linear imaging self scanning (LISS) system. SPOT satellites and IRS series of
satellites extensively use this type of payloads for scanning purposes.
II.91
For multispectral scanning, the along track scanner accommodates one line of detector array
for each spectral band. However, IRS satellites use as many cameras as there are spectral
bands each with a characteristic filter and a linear array of CCD detectors.
The spatial resolution of the sensor = ground resolution cell = GIFOV
The dwell time for the along track scanner is given by
Dwell time = (ground resolution cell dimension) / (velocity of sensor platform)
Under identical setup, it is found that the dwell time of along track scanner is far greater than
the dwell time of the across track scanner. The increased dwell time of the along track
scanner which provides higher signal intensity from ground resolution cells provides two
important improvements for incorporation in the scanner system, namely to minimize the
detector size corresponding to a smaller GIFOV producing higher spatial resolution and to
narrow down the spectral bandwidth, thereby increasing the spectral resolution.
4.5.3 Side Viewing / Side Looking Scanner
The across track and along track scanners described above are used in passive remote sensing
in visible, infrared and microwave regions of the electromagnetic spectrum. These scanners
always receive signals in the nadir view. However, if the user demands (on payment basis) to
observe a dynamic scene frequently, then there is provision in SPOT and IRS satellites to
steer the cameras to look off-nadir at the required scene, some days before to some days after
the normal nadir viewing date.
But scanners of active remote sensing, like the scanners used for radar remote sensing, in
their normal mode of scanning look to the sides and not to the nadir for technical reasons
which will be described later. Therefore such scanners are called side looking airborne radar
(SLAR). The most sought after, sophisticatedly designed synthetic aperture radar (SAR)
belongs to the side looking scanner system.
4.6 Scale, Resolution and Mapping Units
4.6.1 Scale
Remote sensing information is scale dependent. Images can be described in terms of scale
which is determined by the effective focal length of the optics of the remote sensing device,
altitude from which image is acquired and the magnification factor employed in reproducing
the image.
Normally we talk about small scale, intermediate scale and large scale while describing an
image. These scales are quantitatively specified in the following manner:
Thus large scale images provide more detailed information than the small scale images. For
achieving optimum land use planning at different levels the following scales are often
recommended:
II.92
These scales are equally applicable for images of visible, infrared and microwave remote
sensing.
With regard to photo scale (S) of the aerial or satellite imageries, it is computed as the ratio of
the photo distance (d) to the ground distance (D) between any two known points, i.e.
For a photograph taken in the nadir view, the image scale is a function of the focal length of
the camera used to acquire the image, the flying height of the sensor platform above the
ground and the magnification factor (M) employed in reproducing the image, i.e.
II.93
that the object becomes detectable. Because of this, features smaller than 80 m, such as
railway lines and low-order drainage patterns are detectable on Landsat images. Conversely,
on Landsat imagery, objects of medium to low contrast may only be detectable if they are of
the size 250 m or more.
For measurements based on the distinguishability between two point targets, Rayleigh
criterion is used. According to Rayleigh, the two objects are just resolvable if the maximum
(center) of the diffraction pattern of one falls on the first minimum of the diffraction pattern
of the other. Consequently the smallest angle resolvable is
where and D are the wavelength of light and diameter of the lens respectively. In remote
sensing terminology, m is called the instantaneous field of view (IFOV).
Measures of resolution using periodicity of repetitive targets are expressed in line pairs / cm.
Resolution employing spectral properties of the target is the effective resolution element
(ERE). This measure of spectral resolution is of interest because of the increasing importance
of automated classification procedures which are highly dependent upon the fidelity of the
spectral measurements recorded by the sensor system.
Effective resolution element (ERE) is defined as the size of an area for which a single
radiance value can be assigned with reasonable assurance that the response is within 5% of
the value representing the actual relative radiance.
Spatial resolution of a system must be appropriate if one is to discern and analyze the
phenomena of interest, namely the detection and identification of objects and their analysis.
To move from detection to identification, the spatial resolution must improve by about 3
times. To pass from identification to analysis, a further improvement in spatial resolution of
10 or more times may be needed.
4.6.2.2 Spectral Resolution
For a remote sensing instrument, spectral resolution is determined by the bandwidth of the
channels used. High spectral resolution is achieved by narrow bandwidths which are
collectively likely to provide more accurate spectral signature for discrete objects than by
broad bandwidths. However, narrow band instruments tend to acquire data with low signal-
to-noise ratio lowering the system’s radiometric resolution. This problem may be eliminated
if relatively long dwell times are used during scanning / imaging.
4.6.2.3 Radiometric Resolution
Radiometric resolution is determined by the number of discrete levels into which a signal
strength maybe divided (quantization). However, the maximum number of practically
possible quantizing levels for a sensr system depends on the signal-to-noise ratio and the
confidence level that can be assigned when discriminating between the levels.
With a given spectral resolution, increasing the number of quantizing levels or improving the
radiometric resolution will improve discrimination between scene objects. It is found that the
number of quantizing levels has a decided effect upon the ability to resolve spectral radiance
that is related to plant-canopy status. Interdependency between spatial, spectral and
radiometric resolutions for each remote sensing instrument affects the various compromises
and tradeoffs.
II.94
4.6.2.4 Temporal Resolution
Temporal resolution is related to the time interval between two successive visits of a
particular scene by the remote sensing satellite. Smaller the revisit time the better is the
temporal resolution of the sensor system of the remote sensing satellite. High temporal
resolution satellites are more suitable for monitoring dynamic surface features or processes
like the growth and development of agricultural crops, floods and so on. To monitor changes
with time, temporal resolution is also an important consideration when determining the
resolution characteristics of a sensor system. For agricultural applications, the use of time as a
discriminant parameter may allow crops over large areas to be identified with sensors
possessing spatial and spectral resolutions that are too coarse to identify them on the basis of
the spectral and morphological characteristics alone.
4.6.3 Mapping Units
Minimum size of an area which can be recognized over a map is taken as 3mm X 3mm. Thus
this area is taken as the minimum mapping unit. If the scale of the map is 1: 50,000, then the
length of the mapping unit can be calculated as
1 mm on the map = 50,000 mm on the ground
3 mm on the map = 150,000 mm on the ground = 150 m on the ground
II.95
ground truth. In cherry pickers automatic recording sensors can be placed at a height of about
15m from the ground. Towers can be raised for placing the sensors at a greater height for
observation. Towers can be dismantled and moved from place to place. Portable masts
mounted on vehicles can be used to support cameras and other sensors for testing and
collection of reflectance data from field sites.
5.2 Airborne Observation Platform
Airborne observation platforms are important to test the stability performance of the sensors
before they are flown in the space-borne observation platforms. Important airborne
observation platforms include – balloons, drones (short sky spy), aircraft and high altitude
sounding rockets.
5.2.1 Balloon Platform
Balloons are used for remote sensing observations (aerial photography) and nature
conservation studies. Balloons developed by the Socie’te’ Europe’anne de Propulsion can
carry the payloads upto altitudes of 30Km. It consists of a rigid circular base plate for
supporting the entire sensor system which is protected by an insulating and shockproof light
casing. It is roll stabilized and temperature controlled. Essential equipment carried by the
balloon includes – camera, multispectral photometer, power supply units and remote control
system. The sensor system is brought back to earth by tearing the carrying balloon through
remote control.
5.2.2 Airborne High Altitude Photography
Special aircraft carrying large format cameras on vibrationless platforms are traditionally
used to acquire aerial photographs of land surface features. While low altitude aerial
photography results in large scale images providing detailed information on the terrain, the
high altitude, smaller scale images offer advantage to cover a larger study area with a fewer
photographs.
NASA acquired aerial photographs of the United States from U-2 and RB-57 reconnaissance
aircraft at altitudes of 18 Km above the terrain on film of 23 X 23 cm format. Each
photograph covered 839 square kilometers at a scale of 1: 120,000. Black and white, normal
color and infrared color films were used in these missions.
The National High Altitude Photography (NHAP) program (1978), coordinated by the US
Geological Survey, started to acquire coverage of the United States with a uniform scale and
format. From aircraft at an altitude of 12 Km, two cameras (23 X 23 cm format) acquired
black and white and infrared color photographs. The black and white photographs were
acquired with the camera of 152 mm focal length to produce photographs at 1: 80,000 scale
which cover 338 square kilometers per frame. The infrared color photographs were acquired
with a camera of 210 mm focal length to produce photographs at 1: 58,000 scale covering
178 square kilometers per frame.
5.2.3 Airborne Mulitspectral Scanners
Aircraft platforms offer an economical method of testing the remote sensors under
development. Thus photographic cameras, electronic imagers, across-track and along-track
scanners, and radar and microwave scanners have been tested over ground truth sites from
aircraft platforms in many countries, especially in the NASA programmes.
NASA U-2 aircraft acquired images using a Daedalus across-track multispectral (10 bands
between 0.38 – 1.10 m) scanner. Similarly the M-7 airborne scanner having 12 spectral
bands covering ultraviolet, visible and infrared regions, was also tested. Signals from the
II.96
scanners were monitored and controlled in flight at the operator console and were recorded in
analog form by a wide-band magnetic tape recorder. The recorded signals were later digitized
and reformatted on the ground for digital image processing and information extraction.
5.3 Space-borne Observation Platforms
Essentially these are satellite platforms. Two types of satellite platforms are well recognized,
they are: manned satellite platform and unmanned satellite platform.
5.3.1 Manned Satellite Platforms
Manned satellite platforms are used as the last step, for rigorous testing of the remote sensors
on board so that they can be finally incorporated in the unmanned satellites. Crew in the
manned satellites operates the sensors as per the program schedule.
5.3.2 Unmanned Satellite Platforms
Landsat series, SPOT series and IRS series of remote sensing satellites, the NOAA series of
meteorological satellites, the entire constellation of the GPS satellites and the GOES and
INSAT series of geostationary environmental, communication, television broadcast, weather
and earth observation satellites all belong to this unmanned satellite category. We may add to
this list the unmanned satellites launched by Russia, Canada, European Space Agency, China
and Japan to indicate the current state of development in space technology to tackle our
problems from global perspective.
These satellites are space observatories which provide suitable environment in which the
payload can operate, the power to permit it to perform, the means of communicating the
sensor acquired data and spacecraft status to the ground stations, and a capability of receiving
and acting upon commands related to the spacecraft control and operation. The environment
of the space-borne observation platform is considered as both structural and physical and
includes such factors as the framing mount for the payload and functional support systems,
the means of maintaining the thermal levels of the payload within allowable limits, and the
ability to maintain the orbital location of the spacecraft so that the sensors look at their targets
from an acceptable and known perspective. The torso of the observatory is the sensory ring
which mounts most of the observatory subsystems that functionally support the payload. The
satellite mainframe subsystems designed to meet these support functions include: the
structure subsystem, orbit control subsystem, attitude control subsystem, attitude
measurement subsystem, power subsystem, thermal control subsystem, and the telemetry
storage and telecommand subsystems.
5.4 Ground Systems
Ground communication stations are the radio telemetry links between satellites and the earth.
They fall into two general categories: (1) those having the sole function of receiving the
sensor and attitude data from the satellites and (2) those which, in addition to receiving the
sensor and attitude data, can receive satellite house keeping data and transmit commands.
5.5 Satellite Launch Vehicle
Rocket-vehicles are used for launching the satellites. The type of rocket used for Landsat
launches was of Mc Donnell-Douglas Delta-900 series. This was a two-stage rocket that is
thrust-augmented with a modified Thorbooster, using nine Thiokol solid propellant strap-on
engines. The French launch vehicle is an Ariane Rocket. Russian launch vehicles use cryo-
engines. The Indian satellite launch vehicles are of PSLV and GSLV types.
II.97
5.6 Remote Sensing Satellite Orbits
A space-borne remote sensing platform is placed and stabilized (by special orbit maneuvers)
in an orbit in which it moves. From geometrical characteristics point of view, the orbits of of
the space-borne platform can be circular, elliptic, parabolic or hyperbolic. Although the
operational orbits for terrestrial remote sensing are supposed to be circular, it is difficult in
practice to establish and maintain an exactly circular orbit. Therefore, the so-called nominally
circular orbits are slightly elliptical in form. Parabolic and hyperbolic orbits are not used for
terrestrial remote sensing. However, they are used primarily in extraterrestrial flights for
sending us information on the extraterrestrial objects.
From the point of view of periodicity of satellite movement, orbits can be classified as geo-
synchronous (geo-stationary) and sun-synchronous.
5.6.1 Geosynchronous Orbit
It is an important special case of the circular orbit class which is achieved by placing the
satellite at an altitude (35,786, 103 Km) such that it revolves in synchrony with the earth,
namely from west to east at an angular velocity equal to the earth’s rotation rate. The
geosynchronous orbit maintains the satellite over a narrow longitude band over the equator.
When this band shrinks to a line the orbit is called geostationary. These orbits are most
frequently used for communication / television broadcast satellites. They are also used for
meteorological and other applications. Insat series of satellites launched by Indian Space
Research Organization, Department of Space, Government of India belong to this class of
satellites.
5.6.2 Sunsynchronous Orbit
If the orbit precession exactly compensates for the earth’s revolution around the sun, the orbit
is called sunsynchronous. It is an important case of elliptical orbit class in which the orbital
plane is near polar ( 85 degrees from the equatorial plane) and the altitude is such that the
satellite passes over all places on earth having the same latitude twice daily revolving in the
same mode (ascending or descending) at the same local sun time. Here solar incidence angle
which is held almost constant over the samelatitude finds potential applications in earth
resource survey and management. All remote sensing satellites like Landsat, SPOT and IRS
belong to this class of satellites. With sunsynchronous satellites, remote sensing observations
of a particular scene (location) can only be made at one fixed time in nadir view during a
predetermined date which eliminates multitemporal observations within the its revisit period.
5.6.3 Shuttle Orbit
Shuttle orbits range from 200 to 300 kilometers above the surface of the earth at an
inclination of 30o to 60o to the equatorial plane. Payloads are mounted on the cargo bay.
Space shuttles are the manned space observation platforms. Space stations launched by the
United States of America and Russia to carry out important physico-chemical and biological
experiments and to develop some hightech materials are also placed in these orbits so that the
crew can carry food items and ferry the scientists to-and-fro between the earth and the space
station.
5.6.4 Coverage
Coverage represents the areas of the earth which are observed by the satellite in one repeat
cycle. For polar (near polar for technical reasons) orbits, it is given as the maximum north
and south latitudes. An important aspect of the remote sensing mission design of satellite is to
provide coverage of a particular geographic area on some schedule. Coverage has two
elements: (1) the nadir trace or ground track of the satellite, and (2) the sensor view area or
II.98
swath. The ground track is determined by the satellite orbit while the swath is determined not
only by the orbit but also by the field of view and the look direction of the sensor relative to
the nadir. Landsat, SPOT and IRS series of satellites essentially provide coverage through
nadir traces. However, SPOT and the some recent IRS series of satellites have been provided
with tiltable mirrors to sweep off-nadir traces involving programs pointing sequences.
5.6.5 Passes
If one observes the pass sequence of the satellite nadir trace then one finds a daily shift
through N days when on the Nth day the nadir trace over a particular location exactly
coincides. Thus the temporal resolution of the satellite is N days. The characteristic number
of passes and the temporal resolution of different satellite systems are presented in Table 6.
Table 6: Characteristics of some remote sensing satellites
Satellite Altitude Orbits Repetivity Ground Radiometri
(Km) /day ( days) Resolution Resolution
Landsat 918 14 18 79 m 128MSS 4,5,6
1, 2, 3 64 MSS 7
Landsat 705 14.5 16 30 m TM 256 TM
4, 5 120 m TM 6
SPOT 832 26 20 m MS 256
1, 2 10 m PAN
IRS-1A,1B 904 14 22 72.5 m LISS- I 128 LISS-I
36.25m LISS-II 128 LISS-II
IRS-1C 817 24 5.8m PAN 188 m WiFS
II.99
4. American Society of Photogrammetry, (1983). Manual of Remote Sensing. 1, R. N.
Colwell (Ed.), Falls Church, Va.
5. Barbe, D.F. (1980). Charge-Coupled Devices. Topics in Applied Physics. 38.
Springer- Verlag, Berlin.
6. Barzegar, F. (1983). Earth Resources Remote Sensing Platforms. Photogramm. Eng.
Remote Sensing, 49, 1669.
7. Cheng, C.J. (Ed.). (1989). Manual for the Optical-Chemical Processing of Remotely
Sensed Imagery. ESCAP / UNDP Regional Remote Sensing Programme
(RAS/86/141). Beijing, China.
8. Curran, P.J. (1985). Principles of Remote Sensing. Longman, London.
9. Doyal, F.J. (1979). A Large Format Camera for Shuttle. Photogramm. Eng. Remote
Sensing, 45, 73 – 78.
10. Doyle, F.J. (1985). The Large Format Camera on Shuttle Mission 41-G. Photogramm.
Eng. Remote Sensing, 51, 200 – 203.
11. Elachi, C. (1983). Microwave and Infrared Remote Sensors. In: Manual of Remote
Sensing, 1, 2nd ed., R. N. Colwell, (Ed.). American Society of Photogrammetry, Falls
Church, Va. 571 – 650.
12. Elachi, C. (1987). Introduction to the Physics and Techniques of Remote Sensing.
John Wiley & Sons, New York.
13. Fraysse, G. (1989). Platforms. In: Applications of Remote Sensing to
Agrometeorology, F. Toselli, (Ed.). Kluwer Academic Publishers, Norwell, Ma. 35 –
55.
14. Freden, Stanley, C. and Frederick, G. Jr. (1983). Landsat Satellites. In: Manual of
Remote Sensing, Vol. I, 2nd ed., R. N. Colwell, (Ed.). American Society of
Photogrammetry, Falls Church, Va. 517 – 570.
15. Gregory, R.L. (1966). Eye and Brain-The Psychology of Seeing. World University
Library, McGraw-Hill Book Co., New York.
16. Hartl, Ph. (1989). Sensors. In: Applications of Remote Sensing to Agrometeorology,
F. Toselli, (Ed.). Kluwer Academic Publishers, Norwell, Ma. 19 – 34.
17. Jones, R. C. (1968). How Images are Detected. Scientific American, 219, 111 – 117.
18. Joseph, G., Ayengar, V.S., Rattan R., Nagachenchaiah, K. Kiran Kumar, A.S.,
Aradhye, B. V., Gupta, K. K. and Samudraiah, D.R.M. (1996). Cameras for Indian
Remote Sensing Satellite IRS-1C. Current Science (Special Issue), 70 (7), 510 – 515.
19. Kalyanaraman, S. and Rajangam, K. (1996). Advanced Features and Specifications of
IRS-1C Spacecraft. Current Science (Special Issue), 70 (7), 501 – 509.
20. Lillesand, T.M. and Ralph, W.K. (1987). Remote Sensing and Image Interpretation.
2nd ed. John Wiley & Sons, New York.
21. Meisner, D.E. (1986). Fundamentals of Airborne Video Remote Sensing. Remote
Sensing Environ. 19, 63 – 80.
22. Sabins, F., F., Jr. (1987). Remote Sensing - Principles and Interpretations. 2nd ed. W.
H. Freeman and Company, New York.
II.100
23. Shivakumar, S.K., Sarma, K.S., Nagarajan, N. Raykar, M.G., Rao, H.R., Prabhu, K.
V. S. R. and Paramananthan, N.R. (1996). IRS-1C Mission Planning, Analysis and
Operations. Current Science (Special Issue), 70 (7), 516 – 523.
24. Slater, P.N. (1980). Remote Sensing: Optics and Optical Systems. Addition-Wesley.
Reading, Mass.
25. Toselli, F. (Ed.), (1989). Applications of Remote Sensing to Agrometeorology.
Kluwer Academic Publishers. Dordrecht, The Netherlands.
II.101
DIGITAL IMAGE PROCESSING
Anil Rai
I.A.S.R.I., New Delhi -110012
1. Introduction
The roots of remote sensing reach back into ground and aerial photography. But modern
remote sensing really took off as two major technologies evolved more or less
simultaneously: i) the development of sophisticated electro-optical sensors that operate from
air and space platforms and ii) the digitizing of data that were then in the right formats for
processing and analysis by versatile computer-based programs. Today, analysts of remote
sensing data spend much of their time at computer stations, but nevertheless still also use
actual imagery (in photo form) that has been computer-processed.
Image processing procedures fall into three broad categories: Image Restoration
(Preprocessing); Image Enhancement; Classification and Information Extraction. Apart
from preprocessing the techniques of contrast stretching, density slicing, and spatial filtering
will be discussed. Under Information Extraction, ratioing and principal components analysis
have elements of Enhancement but lead to images that can be interpreted directly for
recognition and identification of classes and features.
The data in satellite remote sensing is in the form of Digital Number or DN value. It is said
that the radiances, such as reflectance and emittances, which vary through a continuous range
of values are digitized onboard the spacecraft after initially being measured by the sensor(s)
in use. Ground instrument data can also be digitized at the time of collection. Or, imagery
obtained by conventional photography is capable of digitization. A DN is simply one of a set
of numbers based on powers of 2, such as 26 or 64. The range of radiances, which instrument-
wise, can be, for example, recorded as varying voltages if the sensor signal is one which is,
say, the conversion of photons counted at a specific wavelength or wavelength intervals. The
lower and upper limits of the sensor's response capability form the end members of the DN
range selected. The voltages are divided into equal whole number units based on the
digitizing range selected. Thus, a IRS band can have its voltage values - the maximum and
minimum that can be measured - subdivided into 28 or 256 equal units. These are arbitrarily
set at 0 for the lowest value, so the range is then 0 to 255.
2. Preprocessing
Preprocessing is an important and diverse set of image preparation programs that act to offset
problems with the band data and recalculate DN values that minimize these problems.
Among the programs that optimize these values are atmospheric correction (affecting the
DNs of surface materials because of radiance from the atmosphere itself, involving
attenuation and scattering); sun illumination geometry; surface-induced geometric distortions;
spacecraft velocity and attitude variations (roll, pitch, and yaw); effects of Earth rotation,
elevation, curvature (including skew effects), abnormalities of instrument performance
(irregularities of detector response and scan mode such as variations in mirror oscillations);
loss of specific scan lines (requires destriping), and others. Once performed on the raw data,
these adjustments require appropriate radiometric and geometric corrections.
Resampling is one approach commonly used to produce better estimates of the DN values for
individual pixels. After the various geometric corrections and translations have been applied,
the net effect is that the resulting redistribution of pixels involves their spatial displacements
to new, more accurate relative positions. However, the radiometric values of the displaced
pixels no longer represent the real world values that would be obtained if this new pixel array
could be re-sensed by the scanner (this situation is alleviated somewhat if the sensor is a
Charge-Coupled Device (CCD). The particular mixture of surface objects or materials in the
original pixel has changed somewhat (depending on pixel size, number of classes and their
proportions falling within the pixel, extent of continuation of these features in neighboring
pixels [a pond may fall within one or just a few pixels; a forest can spread over many
contiguous pixels]). In simple words, the corrections have led to a pixel that at the time of
sampling covered ground A being shifted to a position that have A values but should if
properly located represent ground B.
An estimate of the new brightness value (as a DN) that is closer to the B condition is made by
some mathematical re-sampling technique. Three sampling algorithms are commonly used:
In the Nearest Neighbor technique, the transformed pixel takes the value of the closest pixel
in the pre-shifted array. In the Bilinear Interpolation approach, the average of the DNs for the
4 pixels surrounding the transformed output pixel is used. The Cubic Convolution technique
averages the 16 closest input pixels; this usually leads to the sharpest image.
3. False Color Composite
The first example of a color composite, made by combining (either photographically or with
a computer-processing program) any three bands of images with some choice of color filters,
usually blue, green, and red. The customary false color composite made by projecting a green
band image through a blue filter, a red band through green, and the photographic infrared
image through a red filter.
4. True Color View
By projecting IRS Bands 1, 2, and 3 through blue, green, and red filters respectively, a quasi-
true color image of a scene can be generated.
In practice, we use various color mapping algorithms to facilitate visual interpretation of an
image, while analytical treatment usually works with the original DN (digital number) values
of the pixels. The original DN values contain all of the information in the scene and though
their range of values may make it necessary to re-map them to create a good display, it
doesn't add information. In fact, although visual interpretation is easier with the remapped
image, re-mapping loses and distorts information thus, for analytical work, we use the
original DN values or DN values translated to calibrated radiances.
With this mapping, we see a pleasing and satisfying image because it depicts the world in the
general color ranges with which we are naturally familiar. We can imagine how this scene
would appear if we were flying over it at a high altitude.
II.103
5. Other Color Combinations
Other combinations of bands and color filters (or computer assignments) produce not only
colorful new renditions but in some instances bring out or call attention to individual scene
features that, although usually present in more subtle expressions in the more conventional
combinations, now are easier to spot and interpret.
6. Contrast Stretching and Density Slicing
Almost without exception, the image will be significantly improved if one or more of the
functions called Enhancement are applied. Most common of these is contrast stretching. This
systematically expands the range of DN values to the full limits determined by byte size in
the digital data. For IRS this is determined by the eight-bit mode or 0 to 255 DNs. Examples
of types of stretches and the resulting images are shown. Density slicing is also examined.The
two most common image processing routines for improving scene quality fall into the
descriptive category of Image Enhancement or Transformation.The first image enhancer,
contrast stretching, is to enhance their pictorial quality. Different stretching options are
described next, followed by a brief look at density slicing
The contrast stretching, which involves altering the distribution and range of DN values, is
usually the first and commonly a vital step applied to image enhancement. Both casual
viewers and experts normally conclude from direct observation that modifying the range of
light and dark tones (gray levels) in a photo or a computer display is often the single most
informative and revealing operation performed on the scene. When carried out in a photo
darkroom during negative and printing, the process involves shifting the gamma (slope) or
film transfer function of the plot of density versus exposure (H-D curve). This is done by
changing one or more variables in the photographic process, such as, the type of recording
film, paper contrast, developer conditions, etc. Frequently the result is a sharper, more
pleasing picture, but certain information may be lost through trade-offs, because gray levels
are "overdriven" into states that are too light or too dark.
Contrast stretching by computer processing of digital data (DNs) is a common operation,
although we need some user skill in selecting specific techniques and parameters (range
limits). The reassignment of DN values is based on the particular stretch algorithm chosen
(see below). Values are accessed through a Look-Up Table (LUT).
The fundamental concepts that underlie how and why contrast stretching is carried out are
summarized in this diagram: From Lillesand and Kiefer, Remote Sensing and Image
Interpretation, 4th Ed., 1999.
II.104
In the top plot (a), the DN values range from 60 to 158 (out of the limit available of 0 to 255).
But below 108 there are few pixels, so the effective range is 108-158. When displayed
without any expansion (stretch), as shown in plot b, the range of gray levels is mostly
confined to 40 DN values, and the resulting image is of low contrast - rather flat.
In plot c, a linear stretch involves moving the 60 value to 0 and the 158 DN to 255; all
intermediate values are moved (stretched) proportionately. This is the standard linear stretch.
But no accounting of the pixel frequency distribution, shown in the histogram, is made in this
stretch, so that much of the gray level variation is applied to the scarce to absent pixels with
low and high DNs, with the resulting image often not having the best contrast rendition. In d,
pixel frequency is considered in assigning stretch values. The 108-158 DN range is given a
broad stretch to 38 to 255 while the values from DN 107 to 60 are spread differently - this is
the histogram-equalization stretch. In the bottom example, e, some specific range, such as the
infrequent values between 60 and 92, is independently stretched to bring out contrast gray
levels in those image areas that were not specially enhanced in the other stretch types.
Commonly, the distribution of DNs (gray levels) can be uni-modal and may be Gaussian
(distributed normally with a zero mean), although skewing is usual. Multi-modal distributions
(most frequently, bimodal but also poly-modal) result if a scene contains two or more
dominant classes with distinctly different (often narrow) ranges of reflectance. Upper and
lower limits of brightness values typically lie within only a part (30 to 60%) of the total
available range. The (few) values falling outside 1 or 2 standard deviations may usually be
discarded (histogram trimming) without serious loss of prime data. This trimming allows the
new, narrower limits to undergo expansion to the full scale (0-255 for IRS data).
Linear expansion of DN's into the full scale (0-255) is a common option. Other stretching
functions are available for special purposes. These are mostly nonlinear functions that affect
the precise distribution of densities (on film) or gray levels (in monitor image) in different
ways, so that some experimentation may be required to optimize results. Commonly used
special stretches include: 1) Piecewise Linear, 2) Linear with Saturation 3) Logarithmic, 4)
Exponential 5) Ramp Cumulative Distribution Function, 6) Probability Distribution Function,
and 7) Sinusoidal Linear with Saturation.
II.105
7. Spatial Filtering
Just as contrast stretching strives to broaden the image expression of differences in spectral
reflectance by manipulating DN values, so spatial filtering is concerned with expanding
contrasts locally in the spatial domain. Thus, if in the real world there are boundaries between
features on either side of which reflectance (or emissions) are quite different (notable as sharp
or abrupt changes in DN value), these boundaries can be emphasized by any one of several
computer algorithms (or analog optical filters). The resulting images often are quite
distinctive in appearance. Linear features, in particular, such as geologic faults can be made
to stand out. The type of filter used, high- or low-pass, depends on the spatial frequency
distribution of DN values and on what the user wishes to accentuate.
Another processing procedure falling into the enhancement category that often divulges
valuable information of a different nature is spatial filtering. Although less commonly
performed, this technique explores the distribution of pixels of varying brightness over an
image and, especially detects and sharpens boundary discontinuities. These changes in scene
illumination, which are typically gradual rather than abrupt, produce a relation that we
express quantitatively as "spatial frequencies". The spatial frequency is defined as the number
of cycles of change in image DN values per unit distance (e.g., 10 cycles/mm) along a
particular direction in the image. An image with only one spatial frequency consists of
equally spaced stripes (raster lines). For instance, a blank TV screen with the set turned on
has horizontal stripes. This situation corresponds to zero frequency in the horizontal direction
and a high spatial frequency in the vertical.
In general, images of practical interest consist of several dominant spatial frequencies. Fine
detail in an image involves a larger number of changes per unit distance than the gross image
features. The mathematical technique for separating an image into its various spatial
frequency components is called Fourier analysis. After an image is separated into its
components (done as a "Fourier Transform"), it is possible to emphasize certain groups (or
"bands") of frequencies relative to others and recombine the spatial frequencies into an
enhanced image. Algorithms for this purpose are called "filters" because they suppress (de-
emphasize) certain frequencies and pass (emphasize) others. Filters that pass high frequencies
and, hence, emphasize fine detail and edges, are called high pass filters. Low pass filters,
which suppress high frequencies, are useful in smoothing an image, and may reduce or
eliminate "salt and pepper" noise.
Convolution filtering is a common mathematical method of implementing spatial filters. In
this, each pixel value is replaced by the average over a square area centered on that pixel.
Square sizes typically are 3 x 3, 5 x 5, or 9 x 9 pixels but other values are acceptable. As
applied in low pass filtering, this tends to reduce deviations from local averages and thus
smoothes the image. The difference between the input image and the low pass image is the
high pass-filtered output. Generally, spatially filtered images must be contrast stretched to use
the full range of image display. Nevertheless, filtered images tend to appear flat.
8. Ratioing
Ratioing is an enhancement process in which the DN value of one band is divided by that of
any other band in the sensor array. If both values are similar, the resulting quotient is a
number close to 1. If the numerator number is low and denominator high, the quotient
approaches zero. If this is reversed (high numerator; low denominator) the number is well
above 1. These new numbers can be stretched or expanded to produce images with
considerable contrast variation in a black and white rendition. Certain features or materials
can produce distinctive gray tones in certain ratios. Three band ratio images can be combined
II.106
as color composites, which highlight certain features in distinctive colors. Ratio images also
reduce or eliminate the effects of shadowing.
For each pixel, we divide the DN value of any one band by the value of another band. This
quotient yields a new set of numbers that may range from zero (0/1) to 255 (255/1) but the
majority are fractional (decimal) values between 0 and typically 2 - 3 (e.g., 82/51 = 1.6078...;
114/177 = 0.6440...). We can rescale these to provide a gray-tone image, in which we can
reach 16 or 256 levels, depending on the computer display limits. One effect of ratioing is to
eliminate dark shadows, because these have values near zero in all bands, which tends to
produce a "truer" picture of hilly topography in the sense that the shaded areas are now
expressed in tones similar to the sunlight sides.
Three pairs of ratio images can be co-registered (aligned) and projected as color composites.
In individual ratio images and in these composites, certain ground features tend to be
highlighted, based on unusual or anomalous ratio values.
9. Classification
This section deals with the process of classifying multispectral images into patterns of
varying gray or assigned colors that represent either clusters of statistically different sets of
multiband data (radiances expressed by their DN values), some of which can be correlated
with separable classes/features/materials (Unsupervised Classification), or numerical
discriminators composed of these sets of data that have been grouped and specified by
associating each with a particular class, etc. whose identity is known independently and
which has representative areas (training sites) within the image where that class is located
(Supervised Classification). The principles involved in classification, mentioned briefly in
this section. This page also describes the approach to unsupervised classification and gives
examples; it is pointed out that many of the areas classified in the image by their cluster
values may or may not relate to real classes (misclassification is a common problem).
There are two of the common methods for identifying and classifying features in images:
Unsupervised and Supervised Classification. Closely related to Classification is the approach
called Pattern Recognition.
Before starting, it is well to review several basic principles with the aid of this diagram:
II.107
In the upper left are plotted spectral signatures for three general classes: Vegetation; Soil;
Water. The relative spectral responses (reflectance in this spectral interval), in terms of some
unit, e.g., reflected energy in appropriate units or percent (as a ratio of reflected to incident
radiation, times 100), have been sampled at three wavelengths. (The response values are
normally converted [either at the time of acquisition on the ground or aircraft or spacecraft]
to a digital format, the DNs or Digital Numbers cited before, commonly subdivided into units
from 0 to 255 [28]).
For this specific signature set, the values at any two of these wavelengths are plotted on the
upper right. It is evident that there is considerable separation of the resulting value points in
this two-dimensional diagram. In reality, when each class is considered in terms of
geographic distribution and/or specific individual types (such as soybeans versus wheat in the
Vegetation category), as well as other factors, there will be usually notable variation in one or
both chosen wavelengths being sampled. The result is a spread of points in the two-
dimensional diagram (known as a scatter diagram), as seen in the lower left. For any two
classes this scattering of value points may or may not overlap. In the case shown, which treats
three types of vegetation (crops), they don't. The collection of plotted values (points)
associated with each class is known as a cluster. It is possible, using statistics that calculate
means, standard deviations, and certain probability functions, to draw boundaries between
clusters, such that arbitrarily every point plotted in the spectral response space on each side of
a boundary will automatically belong the class or type within that space. This is shown in the
lower right diagram, along with a single point "w" which is an unknown object or pixel (at
some specific location) whose identity is being sought. In this example, w plots just in the
soybean space.
Thus, the principle of classification (by computer image-processing) boils down to this: Any
individual pixel or spatially grouped sets of pixels representing some feature, class, or
II.108
material is characterized by a (generally small) range of DNs for each band monitored by the
remote sensor. The DN values (determined by the radiance averaged over each spectral
interval) are considered to be clustered sets of data in 2-, 3-, and higher dimensional plotting
space. These are analyzed statistically to determine their degree of uniqueness in this spectral
response space and some mathematical function(s) is/are chosen to discriminate the resulting
clusters.
Two methods of classification are commonly used: Unsupervised and Supervised. The logic
or steps involved can be grasped from these flow diagrams:
FORM IMAGES OF
DATA
SEPARATE DATA
INTO GROUPS
WITH CLUSTERING
CHOOSE TRAINING
PIXELS FOR
EACH CATEGORY
CLASSIFY DATA
INTO GROUPS
CALCULATE
STATISTICAL
DISCRIPTIONS
ASSIGN NAME
TO EACH GROUP
NO
SATISFACTORY
?
YES
NO SATISFACTORY
?
CLASSIFY DATA
INTO CATEGORIES
DEFINED
YES
A B
In unsupervised classification any individual pixel is compared to each discrete cluster to see
which one it is closest to. A map of all pixels in the image, classified, as to which cluster each
pixel is most likely to belong, is produced (in black and white or more commonly in colors
assigned to each cluster. This then must be interpreted by the user as to what the color
patterns may mean in terms of classes, etc. that are actually present in the real world scene;
this requires some knowledge of the scene's feature/class/material content from general
experience or personal familiarity with the area imaged. In supervised classification the
interpreter knows beforehand what classes, etc. are present and where each is in one or more
locations within the scene. These are located on the image, areas containing examples of the
class are circumscribed (making them training sites), and the statistical analysis is performed
on the multiband data for each such class. Instead of clusters then, one has class groupings
with appropriate discriminant functions that distinguish each (it is possible that more than one
class will have similar spectral values but unlikely when more than 3 bands are used because
different classes/materials seldom have similar responses over a wide range of wavelengths).
II.109
All pixels in the image lying outside training sites are then compared with the class
discriminants, with each being assigned to the class it is closest to - this makes a map of
established classes (with a few pixels usually remaining unknown) which can be reasonably
accurate (but some classes present may not have been set up; or some pixels are
misclassified.
9.1 Unsupervised Classification
In an unsupervised classification, the objective is to group multiband spectral response
patterns into clusters that are statistically separable. Thus, a small range of digital numbers
(DNs) for, say 3 bands, can establish one cluster that is set apart from a specified range
combination for another cluster (and so forth). Separation will depend on the parameters we
choose to differentiate. We can visualize this process with the aid of this diagram, taken from
Sabins, "Remote Sensing: Principles and Interpretation." 2nd Edition, for four classes:
A = Agriculture; D= Desert; M = Mountains; W = Water.
From F.F. Sabins, Jr., "Remote Sensing: Principles and Interpretation." 2nd Ed.,© 1987.
Reproduced by permission of W.H. Freeman & Co., New York City.
We can modify these clusters, so that their total number can vary arbitrarily. When we do the
separations on a computer, each pixel in an image is assigned to one of the clusters as being
most similar to it in DN combination value. Generally, in an area within an image, multiple
pixels in the same cluster correspond to some (initially unknown) ground feature or class so
that patterns of gray levels result in a new image depicting the spatial distribution of the
clusters. These levels can then be assigned colors to produce a cluster map. The trick then
becomes one of trying to relate the different clusters to meaningful ground categories. We do
this by either being adequately familiar with the major classes expected in the scene, or,
where feasible, by visiting the scene (ground truthing) and visually correlating map patterns
to their ground counterparts. Since the classes are not selected beforehand, this latter method
is called Unsupervised Classification.
The most of the image-processing program employs a simplified approach to Unsupervised
Classification. Input data consist of the DN values of the registered pixels for the 3 bands
used to make any of the color composites. Algorithms calculate the cluster values from these
II.110
bands. It automatically determines the maximum number of clusters by the parameters
selected in the processing. This process typically has the effect of producing so many clusters
that the resulting classified image becomes too cluttered and, thus, more difficult to interpret
in terms of assigned classes. To improve the interpretability, we first tested a simplified
output and thereafter limited the number of classes displayed to 15 (reduced from 28 in the
final cluster tabulation).
9.2. Supervised Classification
The principles behind Supervised Classification are considered in more detail. The fact that
the pixel DNs for a specified number of bands are selected from areas in the scene that are a
priori of known identity, i.e., can be named as classes of real features, materials, etc. allows
establishment of training sites that become the basis of setting up the statistical parameters
used to classify pixels outside these sites.
Supervised classification is much more accurate for mapping classes, but depends heavily on
the cognition and skills of the image specialist. The strategy is simple: the specialist must
recognize conventional classes (real and familiar) or meaningful (but somewhat artificial)
classes in a scene from prior knowledge, such as, personal experience with the region, by
experience with thematic maps, or by on-site visits. This familiarity allows the specialist to
choose and set up discrete classes (thus supervising the selection) and then assign them
category names. The specialists also locate training sites on the image to identify the classes.
Training Sites are areas representing each known land cover category that appear fairly
homogeneous on the image (as determined by similarity in tone or color within shapes
delineating the category). Specialists locate and circumscribe them with polygonal
boundaries drawn (using the computer mouse) on the image display. For each class thus
outlined, mean values and variances of the DNs for each band used to classify them are
calculated from all the pixels enclosed in the site. More than one polygon can be established
for any class. When DNs are plotted as a function of the band sequence (increasing with
wavelength), the result is a spectral signature or spectral response curve for that class. In
reality the spectral signature is for all of the materials within the site that interact with the
incoming radiation. Classification now proceeds by statistical processing in which every pixel
is compared with the various signatures and assigned to the class whose signature comes
closest. A few pixels in a scene do not match and remain unclassified, because these may
belong to a class not recognized or defined.
Many of the classes in general are almost self-evident ocean water, waves, beach, marsh,
shadows. In practice, we could further sequester several such classes. For example, we might
distinguish between ocean and bay waters, but their gross similarities in spectral properties
would probably make separation difficult. Other classes that are likely variants of one
another, such as, slopes that faced the morning sun as IRS flew over versus slopes that face
away, might be warranted. Some classes are broad-based, representing two or more related
surface materials that might be separable at high resolution but are inexactly expressed in the
IRS image. In this category we can include trees, forests, and heavily vegetated areas (the
golf course or cultivated farm fields).
Note that software does not name them during the stage when the signatures are made.
Instead, it numbers them and names are assigned later. Several classes gain their data from
more than one training site. Most of the software has a module that plots the signature of each
class.
II.111
9.2.1 Minimum Distance Classification
One of the simplest supervised classifiers is the parallelopiped method. But a somewhat
better approach (in terms of greater accuracy) known as the Minimum Distance classifier.
This sets up clusters in multidimensional space, each defining a distinct (named) class. Any
pixel is then assigned to that class if it is closest to shortest vector distance.
We initiate our exemplification of Supervised Classification by producing one using the
Minimum Distance routine. The software program acts on DNs in multidimensional band
space to organize the pixels into the classes we choose. Each unknown pixel is then placed in
the class closest to the mean vector in this band space We can elect to combine classes to
have either color themes (similar colors for related classes) and/or to set apart spatially
adjacent classes by using disparate colors
9.2.2 Maximum Likelihood Classification
The most powerful classifier in common use is that of Maximum Likelihood. Based on
statistics (mean; variance/covariance), a (Bayesian) Probability Function is calculated from
the inputs for classes established from training sites. Each pixel is then judged as to the class
to which it most probably belongs. This is done with the IRS data, using three reflected
radiation bands. The result is a pair of quite believable classification maps whose patterns
(the classes) seem to closely depict reality but keep in mind that several classes are not
normal components of the actual ground scene, e.g., shadows.
In many instances the most useful image processing output is a classified scene. This is
because you are entering a partnership with the processing program to add information from
the real world into the image you are viewing, in a systematic way, in which you try to
associate names of real features or objects with the spectral/spatial patterns evident in
individual bands, color composites, or PCI images. The most of the software are capable of
producing both unsupervised and supervised classifications.
References
1. Lillesand, M. and Kiefer, R.W. (1999). Remote Sensing and Image Interpretation, 4th
Ed., John Wiley and Sons, New York.
2. Sabins, F.F. Jr. (1987). Remote Sensing: Principles and Interpretation. 2nd Ed.
Reproduced by permission of Freeman, W.H. and Co., New York City.
II.112
MAP PROJECTIONS AND COORDINATE SYSTEM
Milap Punia
Jawaharlal Nehru University, New Delhi -110067
1. Introduction
The basic problem of map projections is the representation of a curved surface in a plane.
Applicationally, it is often the problem of representing the earth or the moon on a flat map.
The figure of the earth is usually represented by a solid of revolution, either the ellipsoid or
the sphere, as the case may be, which is regarded as a reference surface to which all physical
points are related. These points may be situated on dry land surface of the earth, on the
surface of seas, oceans and lakes or below such water surfaces is the representation of the
mean sea level and its continuation under dry land or over dry land depressions. This
implication is not, strictly speaking, true since the figure of the earth is truly represented only
by an equipotential surface at the mean sea level, called the geoid, and such an equipotential
surface is irregular, or undulating, impossible to express by a rigorous mathematical formula.
The determination of the geoid and the choice of a regularly shaped reference surface which
would best represent the figure of the earth is one of the tasks of geodesy. In recent years this
task has been extended to the moon, and thus, we have a new branch of science dealing with
the size and shape of the moon called selenodesy.
It will, however, be assumed that the currently determined reference surfaces for the earth
and moon are an ellipsoid and a sphere of known parameters respectively, and thus, the
problem of map projections is confined to the representation of the ellipsoidal and spherical
surfaces on a plane.
It may be further stated that there is no perfect solution to the problem, and this may be
readily surmised by trying to apply an orange peel to a flat table surface. To achieve a
continuous contact between the two surfaces, the orange peel would have to be distorted by
stretching, shrinking or tearing and this, while being a gross oversimplification of the
problem of map projections, is nevertheless a good illustration of the impossibility of a
perfect solution. One may argue that the whole problem is totally irrelevant, since it is
possible to limit ourselves to three dimensional models of the ellipsoid or the sphere, either
physically constructed, as in the case of the globe, or mathematically expressed, as done by
geodesists. Theoretically, such an argument is perfectly valid and the desire for a
representation on a plane surface is purely one of convenience. There are many valid reasons
for desiring such a convenience, the obvious ones being that a flat map is easier to produce
and handle than a globe or a scaled portion of a curved surface and the computations on a
plane are much easier than computations on an ellipsoid or a sphere, even in the age of an
electronic computer.
We may, thus, conclude that all representation of curved surfaces on a plain involve
“stretching” or “shrinking” resulting in distortions or “tearing” resulting in interruptions.
Different techniques of representation are applied to achieve representations which posses
certain properties favourable for the specific purpose at hand, and considering the size of the
area to be represented. The technique of representation is commonly called map projection,
although this term is not to be taken literally, since not all representations are produced for
mapping and certainly not all representations are achieved by means of geometric
projections.
2. Purpose of Map Projections
The representation of the earth or lunar surface on a plane may be required for the purpose of
expressing the position of discrete points on the original surface in a plane coordinate system
and the computation of distances and directions within a system of such discrete points. This
is usually of primary interest to the surveyor and mapper, who deals with areas of usually
limited extent within his scope of activity.
Representations of curved surfaces on a plane are also made for the purpose of graphical
presentations which are primary interest to the geographers as aids to the studies or
topography habitation, climatology, vegetation etc. dealing usually with areas of greater
extent.
Three principal cartographic criteria are applied to the evaluation of map projection
properties:
a) Equidistance – correct representation of distances.
b) Conformality (orthomorphism) – correct representation of shapes.
c) Equivalency – correct representation or areas.
The above three criteria are basic and mutually exclusive, other properties being the
secondary nature. Considering this, it should be stressed that there is no ideal representation,
only the best representation for a given purposes.
3. Method of Map Projection
The methods of projection or transformation may be classified as
a) Direct projection from the ellipsoidal to the projection surface.
b) Double projection involving a transformation from ellipsoidal to spherical surface and
then from the spherical to the projection surface.
We have thus, two kinds of datum surface – ellipsoid and sphere. There are three kinds of
projection surfaces – plane, cone and cylinder, the latter two being developable into a plane.
The transformation from datum to projection surface may be geometrical, semi-geometrical
or mathematical in nature. Very few of the transformations are truly perspective projections
in a geometrical sense.
It is convenient to define a map projection as a systematic arrangement of intersecting lines
on a plane, which represent and have a one-to-one correspondence to the meridians and
parallels on the datum surface. The arrangement follows some consistent principle in order to
fulfill certain required conditions. Each set of new conditions results in a different map
projection and thus, potentially there exist an unlimited number of map projections. In
practice, however, the three above mentioned principal cartographic criteria are applied.
4. Classification of Map Projection
The classification of map projections should follow a standard pattern so that any regular
projection (non-conventional) can be described by a set of criteria and conversely a set of
criteria will define a regular projection. Thus, a classification scheme may follow a number
of criteria subdivided into classes and varieties.
Classes may be considered as different points of view. These points of view are not mutually
exclusive. An example of such non-exclusiveness is the consideration of a person as, say, a
human being, Frenchman, red haired, Protestant, vegetarian etc. This corresponds
II.114
respectively to classes of species, nationality, hair color, religion, type of food consumed etc.
none of which are mutually exclusive.
Varieties are the accepted or existing subdivisions within each class and they are mutually
exclusive since a person would have a certain one nationality, one hair colour, one religion
etc. To facilitate the setting up of a classification scheme for map projections composed of
classes and verities certain specific factors should be considered.
a) The projected object or datum surface.
b) The projection surface upon which the datum surface is projected.
c) The projection or representation per se.
The projection surface is considered as the extrinsic problem and the process of projection or
representation as the intrinsic problem.
4.1 Extrinsic Problem
This problem involves the consideration of the properties of the projection surface relative to
the datum surface giving rise to three classes.
i. Nature of the projection surface defined as geometric figure.
ii. Coincidence or contact of the projection surface with the datum surface.
iii. Position or alignment of the projection surface with relation to the datum surface.
Class I may be further subdivided into three varieties, such representing one of the basic
projection surface, namely, the plane, the cone and the cylinder. The simplest of these
projection surfaces is the plane, which when tangent to the datum surface would have a single
point of contact, this being also the centre of the area of minimum distortion. The cone and
the cylinder, which are both developable into a plane were introduced with view to increase
the extent of contact and consequently the area of minimum distortion.
Class II is thus further subdivided into three varieties representing the three types of
coincidence between the datum and projection surface, namely, tangent, secant and
polysuperficial. It may easily be surmised that tangency between the datum and projection
surfaces results in a point or line contact, the former in the case of projection surface being a
plane and the latter in the case of projection surface being either a cone or a cylinder. To
increase the contact between the surfaces and thus, also the area of minimum distortion, the
secant case has been introduced resulting in two lines of contact when the projection is either
a cone or a cylinder. A still further increase of contact and consequently in areas of minimum
distortion is achieved by employment of polysuperficiality, or in other works, a series of
successive projection surfaces. A series of successive planes would produce a polyhedric
(multiple plane) projection, a series of cones produces a polyconic and a series of cylinders a
polycylinderical projection.
Class III is subdivided into three varieties representing the three basic positions or
alignments of the projection surface relative to the datum of surface, namely, normal,
transverse and oblique. If the purpose of the projection is to represent a limited are of the
datum surface, it is advantageous to achieve the minimum of distortion in that particular area.
This is possible through varying the attitude of the projection surface. If the axis of symmetry
of the projection surface coincides with the rotational axis of the ellipsoid or the sphere, the
normal case is obtained. With the axis of symmetry perpendicular to the axis of rotation, we
have the transverse case and the other attitudes of the axis of symmetry result in oblique
cases.
II.115
4.2 Intrinsic Problem
This problem involves the consideration of a projection from the point of view of its
cartographic properties and the mode of generation, giving rise to the following two classes :
i. Properties
ii. Generation
Class IV is further subdivided into 3 mutually exclusive varieties representing the three basic
cartographic criteria according to which properties are evaluated : equidistance, conformality
(orthomorphism) and equivalency (equality of areas).
Equidistance means that there is a correct representation of a distance between two points on
the datum and the corresponding points on the projection surface, so that the scale is
maintained along the lines connecting a pair of points. This is, however, limited to certain
specified points and is by no means a general property between points on the two surfaces.
Equivalence of areas means of figures represented are retained, but at the expense of shapes
and angles which are in such case deformed.
Conformality means retention of shape or form, and thus also retention of angles (directions),
this property being limited to differentially close points and certainly not valid for areas of
significant dimensions.
Class V is further subdivided into 3 mutually exclusive varieties representing the three
principal modes of generation of projections or representations. The necessity or the desire to
obtain certain properties in projections or representations led to the evolution of the various
modes of generation either through a purely geometric or perspective projection technique or
through a process only partly projective where only one family of lines is projected by a
pencil of planes all passing through the axis of symmetry of the projection surface. There are
some representations, however, which are completely free of the projection operation, i.e. no
rays are involved and the representation is achieved by a convention followed by a purely
mathematical process. There are thus in Class V three varieties, namely, geometric, semi-
geometric and conventional.
It is important to stress the fact that the number of conventional representations is infinite.
There representations are subject to empirical or posterior classifications only since in this
case we are unable to answer questions pertaining to the geometric form of the projection
surface or its position relative to the datum surface.
4.3 The Classification Scheme
The previously discussed classes and varieties may be arranged as follows :
Classes Varieties
II.116
A regular (non0conventional) projection may be described by a set of varieties, one from
each class; and conversely, a set of varieties, one from each class, defines a regular
projection.
4.4 Development based Classification
The best way has been found to conceive a surface that can touch the earth in a point or a line
and the same could be spread flat without distortions. For example, a cone can be put like
cutting them straight along height they can be spread flat – a cylinder like a rectangle and
cone like a sector.
Such surfaces are called developable surfaces. A plane can come in contact on a point of the
globe. The mesh work of latitude and longitude can come on it from any view point, centre of
the earth, on the globe or far away in space.
Thus, we have the following names of projections.
i) Azimuth or zenithal – on a plane Gnomonic, stereographic, orthographic, plane touching
on a point.
ii) Conical – one or two standards, polyconic i.e. every parallel taken standard, Alber’s
Bonne’s cone touching / cutting globe.
iii) Cylindrical – Cassini’s; simple cylindrical, Mercator’s
iv) Borrowed properties – in conventional like Sanson-Flamsteed or sinusoidal; Molliweide,
globular.
5. Map Reference Coordinate System
There are many coordinate system for map reference. These are:
5.1 Systems of Map Numbering
The main system of numbering used on maps published by the Survey of India is on spherical
layouts. These are: (a) the International system and (b) the India and Adjacent Countries
Series system. Another international system of numbering on spherical layout is the World
Geographic Reference (GEOREF) system.
5.2 The International System
The International system only applies strictly to I/M or ¼ inch (1:250,000) sheets and is used
for the International Map of the World (IMW) on the Millionth scale. It is based on a series
of I/M sheet number of this series each covering 40 latitude by 6o longitude. Each sheet
number of this series consists of two letters and a number and defines the geographical
position of the sheet.
The first letter is either “N” or “S” depending on whether the sheet is north or south of the
equator. This is followed by a letter and a number this letter indicates the latitude of the sheet
and is applied in alphabetical order for every 4 belt of latitude from the equator. Thus all
sheets between 0o and 4o latitude are NA and those between 7o and 6o E will be NA-31. The
number of I/M sheet covering Dehradun between latitude 28o and 79o E is NH-44 G. The
international system of numbering does not extent to sheet larger than ¼ inch (1:250,000)
scale.
5.3 The Indian and Adjacent Countries (I.A.C.) System
This is a purely arbitrary system of sheet numbering based on the 4o x 4o layout of 1/M sheets
adopted in India and adjacent countries. The sheet number does not give any indication of
the geographical position of the sheet.
II.117
Each sheet covers an area of 4o latitude by 4o longitude and the numbering is from north to
south increasing from west to east. This series does not extend north of latitude 40o N and
begins with sheet No. 1 with its north-west corner at latitude 40o and longitude 44o E.
Each 1/M sheet of this series is divided into 16 degrees sheets on ¼ inch or 1:250,000 scale.
These are lettered from north to south from A in the north-west corner to P in the south-each,
example, the sheet covering Delhi is 53 H/2 and for Dehradun 53J/15.
Each degree sheet of this series is divided into 16 one inch or 1:50,000 scale sheets each of
15 minutes of latitude and 15 minutes of longitude. These are numbered from 1 in the north-
west corner to 16 in the south-each, example 53 J/15.
Each one-inch or 1:50,000 scale sheet is further sub-divided into 6 sheets on 1:25,000 scale
each of 7-1/2 minutes of latitude and 7-1/2 minutes of longitude. These are numbered as NW
for the north-west corner NE for north-east, SW for south-west and SE for south-east corner
into 4 sheets, for example 53 J/15 NE, 53J/15/NW, 53J/15/SE & 53J/15/SW.
5.4 Universal Transverse Mercator (UTM) System
Universal Transversal Mercator projection is now widely used projection for topographical as
well as large scale mapping purposes.
1) It is conformal cylindrical projection.
2) It covers almost entire world (80o S to 84o N) in 60 zones of 6o Longitude.
3) Reference Latitude (o), is the equator. Each zone has its own Central Meridian.
4) Each zone has its own Cartesian coordinate system.
II.118
References
1. Dana, H. (1995). Map Projections. URL The Geographer's Craft Project. Dept. of
Geography, University of Texas at Austin. Austin.
2. Gersmehl, P. J. (1991). The Language of Maps. Pathways in Geography Series, title
no. 1. Indiana, Pennsylvania: Indiana University of Pennsylvania.
3. Pearson II, F. (1990). Map Projection: Theory and Applications. Boca Raton, Florida:
CRC Press, Inc.
4. Raisz, E. (1962). Principles of Cartography. New York: McGraw-Hill Book
Company.
5. Snyder,P. and Voxland, M. (1989). An Album of Map Projections. U.S. Geological
Survey Professional Paper 1453. Denver: United States Government Printing Office.
6. Maling, D.H. (1973). Coordinate Systems and Map Projections, George Phillip and
Son Limited, London.
II.119
HYPERSPECTRAL REMOTE SENSING
R.N. Sahoo
I.A.R.I., New Delhi -110012
1. Introduction
The term hyperspectral is used to refer to spectra consisting of large number of narrow,
contiguously spaced spectral bands. In the field of remote sensing, the term hyperspectral is
also used interchangeably with other terms such as spectroscopy, spectrometry,
spectroradiometry and rarely ultraspectral. Spectroscopy is a branch of physics concerned
with the production, transmission, measurement and interpretation of electromagnetic spectra.
Spectrometry or spectroradiometry is the measure of photons as a function of wavelength.
Ultraspectral is beyond hyperspectral, a goal that has not been achieved yet. Spectrometers
are used in laboratories, field, aircraft or satellites to measure the reflectance spectra of
natural surfaces. When an image is constructed from an imaging spectrometer that records the
spectra for contiguous image pixels, the terms shift to become imaging spectroscopy, imaging
spectrometry or hyperspectral imaging. Hyperspectral imaging is a new technique for
obtaining a spectrum in each position of a large array of spatial positions so that any one
spectral wavelength can be used to make a recognizable image (Clark, 1999). By analyzing
the spectral features in each pixel, and thus specific chemical bonds in materials, we can
spatially map materials. The narrow spectral channels that constitute hyperspectral sensors
enable the detection of small spectral features that might otherwise be masked within the
broader bands of multi-spectral scanner systems. In this regard, we hypothesis that
hyperspectral sensors could help to overcome the traditional problems faced when using the
broader bands of multi- spectral scanner systems, such as the saturation problem in estimating
quantity and the estimation of quality.
1.1 Advantages of Hyperspectral Remote Sensing
Direct field techniques for estimating soil and vegetation attributes require frequent
destructive sampling. Such techniques are difficult, extremely labour intensive and costly in
terms of time and money. They can hardly be extended to cover large areas. However, remote
sensing technique particularly of high spectral resolution has been found to be very potential
for quantitative assessment of soil and vegetation at spatial scale. A major limitation of
broadband remote sensing products is that they use average spectral information over
broadband widths resulting in loss of critical information available in specific narrow bands
(Blackburn, 1998, Thenkabail et al.. 2000). Recent developments in hyperspectral remote
sensing or imaging spectrometry have provided additional bands within the visible, NIR and
shortwave infrared (SWIR) (Figure 1). Most hyperspectral sensors acquire radiance
information in less than 10 nm bandwidths from the visible to the SWIR (400-2500 nm)
(Asner, 1998). For example, the spectral shift of the red-edge (670-780 nm) slope associated
with leaf chlorophyll content, phenological state and vegetation stress, is not accessible with
broadband sensors (Collins, 1978; Horler, et al.1983).
Figure 1. Comparison of Hyperspectral and multispectral data of a vegetation.
2. Hyperspectral Remote Sensing of Soil
Over the last couple of decades, it has been a challenge to find the most appropriate technique
for studying soil properties effectively and, at the same time, reducing the time and effort
involved in field sampling and laboratory analysis. This has been a major concern not only
for soil scientists but also for environmental specialists. The scope of remote sensing data
has been widely studied, in this respect, looking more closely at how the spectral response of
soil can be linked to various soil properties and characteristics. The relatively high spectral
resolutions as well as contiguous placement of bands covering wider region of electro-
magnetic spectrum provide more opportunities for soil characterization.
2.1 Hyperspectral Reflectance Spectra for soil characterization
Reflectance spectra have been used for many years as one of the source of information about
the variation in the Earth's surface composition (Van der Meer et al. 2001). Consequently, the
soil properties derived from these spectra have been studied as early as 1980s. Study of soil
properties controlling soil reflectance and conversely observations of soil reflectance
provides information on some of the soil properties and, in general, the state of soil quality
(Irons et al. 1989). Many studies have drawn their main emphasis on the visible and infrared
regions of the electro-magnetic spectrum- between 300 and 2500 nm (Baumgardner et
al.1985).
In general, a wide range of information can be obtained from reflectance properties related to
the nature and chemical composition of the soil material (Stoner and Baumgardner, 1981;
Baumgardner et al. 1985). This is mainly based on specific absorption of spectrally active
groups (known as chromophores), such as Fe, OH- in water and minerals, CO32-, Al2+, (Mg+)
OH, SO42- in minerals, and any others in organic matter (Ben-Dor et al. 1997). Reflectance
data have been successfully used for prediction of numerous soil properties, such as soil
moisture and organic matter content in laboratory studies (Dalal and Henry, 1986; Ben-Dor
and Banin, 1995a). Further, many researchers have studied the effect different factors
affecting soil reflectance and thereby could predict them from soil reflectance.
Spectral reflectance has been widely used by many in the visible and near infrared (VIS/NIR,
400- 2500 nm) as well as in mid infrared (MIR) spectral regions for assessment of soil
chemical fertility. Most common fertility parameters studied were Soil organic carbon
(SOC), inorganic carbon (SIC), total nitrogen (TN), cation exchange capacity (CEC), pH,
potassium (K), magnesium (Mg), calcium (Ca), zinc (Zn), iron (Fe), and manganese (Mn)
with various levels of prediction accuracy (e.g. Baumgardner et al. 1985; Ben-Dor and
Bannin, 1995a; Malley et al. 1999; Shepherd and Walsh, 2002; Viscarra Rossel et al. 2006).
II.121
For characterizing soils from hyperspectral reflectance data Kumar et al. (2006) collected
eighty seven surface soil samples (0-15cm) representing 23 soil series were collected from 23
sites of five different places of India encompassing 5 different climatic zones including soils
from 4 soil taxonomic orders (Vertisols from Nagpur, Alfisols from Ranchi and Inceptisols
from IARI farm) on the basis of reports on soils of different states published by the National
Bureau of Soil Survey and Land Use Planning and spectral studies were carried out using FS 3
ASD spectroradiometer (having spectral range 350-2500nm). Reflectance spectra of soil
samples that represent four soil orders within the examined population are given in Fig.1.
These signatures can easily be attributed to several common soil constituents. The most likely
component in the NIR region is the OH groups in both the adsorbed water (at 1.4µm and
1.9µm) and the crystal water of clay minerals (at 1.45µm and 2.2µm ) (Hunt and Salisbury
1970). Also the CO3 in the calcite mineral is very active in the NIR region of soils (major
peak at 2.33µ) (Hunt et al.1971; Gaffey 1986).
From the analysis of the reflectance spectra pattern different soils orders namely Alfisols,
Inceptisols, Mollisols and Vertisols could also be discriminated. The spectral pattern of
Alfisols showed higher reflectance pattern in the IR region (Fig 1) than the others three soils
orders (Inceptisols, Vertisols, and Mollisols), due to high free iron oxide content as well as
due to high kaolinite and interstratified kaolinite content. Alfisols showed high absorption
pattern in the visible region than the Inceptisols and Vertisols. These soils showed strong
absorption peaks near 450 nm and weak absorption peak near 550 nm. The absorption band
near 450 nm is caused by paired and single Fe3+ electron transitions to higher energy state
(Sherman and Waite,1985), the small absorption peak near 550 nm may be due to the
chromophore FeO-OH found in goethite (Mortimore et al. 2004) which is dominated in
Alfisols. In VIS region, convex curve was found which is characteristic of Alfisols soil
orders. These convex curves were broader for high iron and free iron oxide containing soil
samples. Mollisols spectral reflectance was lowest one as compared to all other soil orders
throughout the spectral region (350-2500 nm). Lowest spectral reflectance pattern of
Mollisols was mainly due to the high organic matter content. Spectral reflectance pattern of
Vertisols showed higher reflectance pattern than Mollisols but lower than Inceptisols and
Alfisols. In VIS region concave curve was observed for Vertisols as well as for Mollisols.
The concave and convex pattern of the reflectance spectra were observed in the VIS region
because these were guided with organic matter and iron content of the soils which were
highly correlated in this region. Spectral reflectance of Inceptisols was observed to be higher
than Vertisols and Mollisols but less than Alfisols. It may be due to its low organic matter
content and sandy loam texture.
VIS (350-710 nm) region of spectra was dominated with several features and specific pattern
with respective bands where but NIR (710-1110 nm) region was featureless. (Fig 1). These
four soil orders namely Mollisols, Vertisols, inceptisols, and Alfisols in red bad (650-700 nm)
showed around 10, 15, 25, and more than 25 % reflectance respectively. So VIS region was
found to be more useful for discriminating these four soil orders as compared to NIR region
of spectra. The spectral pattern in short wave infrared (SWIR, 1100-2500 nm) region of
Alfisols shows the characteristic absorption peak near 1400, 2200, and 2300 nm bands. This
1400 nm peak may be attributed due to the first overtone of the O-H stretch and a second
doublet at 2200 and 2300nm is due to the combination Al-OH bend plus O-H stretch. This
metal-OH bend plus O-H stretch combination near 2200 nm and 2300 nm are diagnostic
absorption features for clay minerals (Clark et al. 1990). Inceptisols spectra in SWIR region
also showed characteristic absorption peaks at near 1400, 2200, and 2300 nm bands. These
two soil orders alongwith its characteristics absorption peaks, also have universal water
absorption band near 1900 nm. Alfisols showed stronger reflectance peaks near 2300 nm, due
II.122
to the presence of surface O-H group as compared to Inceptisols. Inceptisols showed
absorption band near 2200 and 23000 due to additional Al-OH features. Vertisols spectra in
SWIR region showed less pronounced peak near 1400 nm band which is also characteristic
absorption peak for Inceptisols and Alfisols. However Vertisols have characteristic broader
and deeper absorption peaks near 1900 nm band. These characteristics absorption peaks was
observed due to presence of smectite clay mineral where water molecules are absorbed
between the sheets of the smectite structure. Clark et al. 1990 have found that the intense
absorption peaks present in smectite dominated soils is due to the combination of the H-O-H
bend with the O-H stretches. Less pronounced absorption peak was also observed near
2200nm. The spectrum that has a 1400 nm band but not 1900 nm band indicates that only O-
H group is present and only a small amount of water, because of a weak 1900 nm absorption.
But presence of a intermediate O-H absorption feature at 1400nm is indicative of Inceptisols.
Vertisols were found to have the least O-H absorbtion peak at 1400 nm. It was observed that,
when the amount of kaolinite content decreases, the characteristic absorption bands at 1400
nm and 2200 nm become less pronounced in Alfisols. Conversely as the amount of smectite
increases in soil samples, the characteristic 1900 nm water absorption band becomes more
broader in vertisols. Spectral pattern of Mollisols in SWIR region were observed to be quite
different than rest three soil orders. Mollisols spectra showed absorption band near 1400 nm
and 1900 nm, which was neither intense like Inceptsols and Alfisols nor broader like
Vertiisols. These may be due to lack of surface O-H group and presence of non expanding
type of minerals (Vermiculite, Mica, Chlorite etc).
(a) (b)
(c) (d)
Figure 1(A). Spectral reflectance pattern of representative soil of four soil orders in (a)
full range, 350-2500nm, (b) only visible range (c) only NIR range and (d)
SWIR range.
II.123
2.2 Quantitative Estimation of Soil Parameters from Hyperspectral Remote Sensing
Hyperspectral data being of larger volume, overlapping of weak overtones and fundamental
vibrational bands, have been very difficult for its direct interpretation (Wetzel, 1983).
Therefore, multivariate analysis is required for quantitative interpretation of soil parameters
from hyperspectral reflectance data. A number of different calibration techniques are
available and have been applied when relating measured spectra to measured values of soil
properties. The choice of calibration technique will depend on the application of the data.
Principal components regression (PCR) and stepwise multiple-linear regression are the most
common (Wise et al. 2003). Different pre-processing transformations have been applied in
numerous studies to transform soil spectral data, remove noise, accentuate features, and
prepare them for chemometric modelling. Pre-processing transformations of spectral data
constitute an important step in multivariate calibration and have been shown to improve the
accuracy of prediction models (Dunn et al. 2002; McCarty et al. 2002).
2.2.1 Multivariate regression approaches
Linear Regression estimates the coefficients of the linear equation, involving one or more
independent variables that could best predict the value of the dependent variable (Gomez and
Gomez, 1984). The complexity and multicollinearity are two common properties inherent to
hyperspectral data. On the other hand, in view of weak correlations of individual reflectance
data with most of the soil properties, univariate modelling is rarely applicable for
hyperspectral use of soil properties. Dardenne (1996) used two multivariate regression
techniques namely multiple linear regression (MLR) and partial least-square regression
(PLSR) to estimate soil parameters from the hyperspectral image data.
Kumar et al. (2009) evaluated the ability of ASD hyperspectral data for estimating 17 soil
properties such as mineralizable nitrogen(N), available phosphorous (P) and potassium (K),
DTPA extractable manganese (Mn), iron (Fe), copper (Cu) and zinc (Zn), calcium carbonate
(CaCO3), soil organic carbon (SOC), pH (1:2.5), EC (1:2.5), bulk density (BD), particle
density (PD), hydraulic conductivity (Ks) and soil texture. The spectral data of 85 soils of
Jalandhar district of Punjab was recorded both at field and laboratory conditions (Fig. 2).
Spectra sensitivity analysis was done to find out best spectral range for different soil
parameter. Example soil organic carbon is given in Fig. 3. Step wise regression approach
was used and results revealed that models developed using derivative spectral data were able
to predict some selected parameters with reasonable higher accuracy while reflectance and
absorbance values did yield poor results. Based on adjustable R2 of the predicted models, first
derivative of absorbance was found suitable for nitrogen while its second derivative was best
for Mn, Fe, and Zn for prediction model. Second derivative of reflectance was selected for P
and Cu and First derivative of reflectance was better for K prediction. The highest
predictability (adjusted R2) was 0.93 recorded for CaCO3 while lowest 0.68 was for N. The
comparison of predicted and measured values of some soil parameters are given in Fig. 4.
II.124
Figure 2. Study area and soil spectral data collection in the farmers’ field in situ and
laboratory condition.
II.125
Figure 4. Evaluation of Predicted models for some soil properties
II.126
2.2.3 Continuum removal
The continuum removal approach aims at quantifying the absorption by materials at a
specific wavelength assuming that no other material has strong absorption features around
this specific wavelength (Clark and Roush, 1984). The continuum is approximated by a
straight line joining the two local reflectance maxima placed on both shoulders (λmin and λmax)
of the peak absorption wavelength (λpeak). Continuum removal, CRλ, was thus written as a
function of reflectance values R(λ) at wavelength λ, with the constraint that its maximum
value could not be above 1.0 (concavity of the reflectance spectra at this location). Specific
absorption peaks for different chemical components mixed with soil can be applied for
variability assessment (Chabrillat et al. 2002).
Santral et al. (2009) could retrieve spectral transformation functions (STF) from measured
soil reflectance data to predict some of most difficult soil parameters e.g. Van Genuchten
Parameters (α and n) which is used to derive soil hydraulic conductivity. The spectral data
was collected using FS3 ASD spectroradiometer and integrated band reflectance
corresponding to ETM+ of Landsat-7, computed CR factor were used to develop the STFs
and predict soil hydraulic parameters. Comparison predicted and observed of value of soil
hydraulic parameters are given in Figure 5.
Figure 5. Observed and predicted values of saturated hydraulic conductivity [ln(Ks)] (a and
b), and van Genuchten parameters, ln(α) (c and d), and van Genuchten parameter,
n (e and f) with the best performances (left side) and worst performances (right
side) among pedotransfer functions and spectrotransfer functions developed with
different combination of predictor variables for each hydraulic property.
II.127
2.3 Quantification of soil properties from imaging spectroscopy
Hyperspectral data analysis can be divided into two categories (Aspinall et al. 2002). One is
top-down, which essentially uses field maps to train the imagery so as to detect certain
features of the examined objects. Field survey and geo-referencing are necessary because
high positioning accuracy is needed in this approach. Atmospheric correction of the imagery
may not be needed. Since the high spatial resolution image usually does not have large extent
coverage, the atmospheric variation within each scene is not noticeable. Associating the
ground features directly with the image enables the classifying algorithm to incorporate
atmospheric effects into the feature spectra to search for similar features on the same image.
However, this approach was not found feasible for large physical extent analysis (Aspinall et
al. 2002). The other approach, in contrast, is bottom-up which typically uses ground- or
laboratory-measured spectral libraries to identify features in the image. Atmospheric
correction is essential because atmospheric effects need to be removed from the image before
it can be quantitatively compared with ground-measured spectra. Geo-referencing and
registration are necessary to match geographic positions of image pixels to ground sites. To
reduce spectral noise, the absorption regions affected by water and carbon-dioxide of the
spectrum should be removed before further processing (Kruse, 1994).
Kadupitiya et al. (2009) evaluated Hyperion sensor of EO-1 for estimation of eight soil
properties such as soil organic content (SOC), CaCO3, Mineralizable Nitrogen (N), available
Phosphorous (P), Potassium (K), sand %, silt % and clay % in the Jalandhar district of Punjab
and compared with the proximal reflectance data (through ASD FS3) collected synchronizing
with the satellite pass. The study showed that, the Hyperion spectral pattern were comparable
with ground measured spectra after atmospheric correction using FLAASH (Fig. 6). Derived
spectral parameter such as first and second derivative of reflectance and absorbance were
found to better suited than original reflectance data for developing prediction models for soil
properties irrespective of the platform of the sensor. Comparison of hyperion data with
proximal reflectance data of spectroradiometer revealed that R2 of prediction models
decreases when we move from using laboratory spectra to field and then to Hyperion sensor.
However, some parameters like SOC and CaCO3 could be predicted with reasonable
accuracy from Hyperion sensor (Fig. 6).
II.128
Figure 6. Comparison of Hyperion reflectance data with field and laboratory reflectance
spectra. Brown and green line is soil reflectance collected in laboratory and field
conditions respectively and blue line for hyperion data. Light blue strip in the
graphs indicates the spectral range not considered due to noise.
II.129
3. Hyperspectral Remote Sensing of Vegetation
Leaves represent the main surfaces of plant canopies where energy and gas are exchanged.
Hence, knowledge of their optical properties is essential to understand the transport of
photons within vegetation (Despan and Jacquemound, 2004). The general shape of
reflectance and transmittance curves for green leaves is similar for all species. It is controlled
by absorption features of specific molecules and the cellular structure of the leaf tissue (Ustin
et al.1999). Three spectral domains can be distinguished (Fig 8): In the visible domain (400-
700 nm) absorption by leaf pigments is the most important process leading to low reflectance
and transmittance values. The main light absorbing pigments are chlorophyll a and b,
carotenoids, xanthophylls, and polyphenols. Chlorophyll a is the major pigment of higher
plants and together with chlorophyll b account for 65 percent of the total pigments.
Chlorophyll a and b have absorption bands in the blue at around 430/450 nm and in the red
domain at around 660/640 nm. These strong absorption bands induce a reflectance peak in
the green domain at about 550 nm. Carotenoids and xanthophylls absorb mainly in the blue
and are responsible for the colour of flowers, fruits, and the yellow colour of leaves in
autumn. Polyphenols (brown pigments) absorb with decreasing intensity from the blue to the
red and appear when the leaf is dead (Verdebout et al. 1994).
In the near-infrared domain (near-IR: 700-1300 nm) leaf pigments and cellulose are almost
transparent, so that absorption is very low and reflectance and transmittance reach their
maximum values. The level of reflectance on the near-IR plateau increases with increasing
number of intercell spaces, cell layers, and cell size. Scattering occurs mainly due to multiple
refractions and reflections at the boundary between hydrated cellular walls and air spaces
(Guyot, 1990).
In the mid-infrared domain (mid-IR: 1300-2500 nm), also called shortwave-infrared (SWIR),
leaf optical properties are mainly affected by water and other foliar constituents. The major
water absorption bands occur at 1450, 1940, and 2700 nm and secondary features at 960,
1120, 1540, 1670, and 2200 nm (Ustin et al. 1999). Water largely influences the overall
reflectance in the mid-IR domain and also has an indirect effect on the visible and near-IR
reflectances. Protein, cellulose, lignin, and starch also influence leaf reflectance in the mid-
IR. However, the absorption peaks of those organic substances are rather weak as they result
from overtones or combinations related to fundamental molecular absorptions in the region 5
to 8 µm (Curran, 1989; Table 1). The molecular absorptions are associated with certain
chemical bonds, such as C-H, N-H, C-O, and O-H. In fresh leaves, spectral features related to
organic substances are masked by the leaf water, so that estimation of leaf constituents is
difficult (Verdebout et al.1994).
II.130
Figure 8. Typical Reflectance Curve of Vegetation and causes of spectral characteristics
(Jensen, 2000).
II.131
Table 1: Absorption features in vegetation reflectance spectra related to leaf nitrogen
concentration (Adapted from Curran, 1989; Lucas and Curran, 1999)
The spectral properties of live foliage set up the radiation field in a canopy, and these spectral
properties express the presence and abundance of both the inputs and products of
photosynthesis (Figure 9). Leaf pigments such as chlorophylls, carotenoids, and anthocyanins
are expressed in the 400-700 nm range, matching the wavelength region of maximum solar
input to the biosphere. The relative contribution of pigments to the reflectance and
transmittance properties of foliage varies by wavelength in this region, and all pigments have
overlapping absorption features (Figure 10). Chlorophyll a (chl-a) displays maximum
absorptions in the 410- 430 nm and 600-690 nm regions; whereas chlorophyll b (chl-b) shows
maximum absorptions in the 450-470 nm range. Carotenoids absorb most efficiently between
440 and 480 nm. In the foliage of many canopy species, chl-b dominates the overall
absorption spectrum at shorter and longer wavelengths in the visible spectrum, whereas
carotenoids can be a major contributor at slightly longer wavelengths (gray line, Figure 10).
In the near-infrared (700-1300 nm) and shortwave-infrared (1300-2500 nm), the leaf
spectrum is dominated by water content, thickness, and, to a lesser degree, protein-nitrogen
(N), cellulose and lignin content (Figure 9) (Curran 1989). In particular, live foliage is a very
efficient at scattering radiation in the 750-1300 nm range; this is caused by internal scattering
at the air-cell-water interfaces within the leaves (Thomas et al.. 1971, Hunt et al.. 1987). At
longer wavelengths (> 1300 nm), relatively small amounts of water (and resulting air-water
interfaces) 4 effectively trap radiation, resulting in absorption that exceeds scattering
processes in this portion of the spectrum. Meanwhile, leaf structural properties, especially
area per mass (specific leaf area or SLA), play an underlying but integral role in containing
the biochemicals in a functional form adapted for leaf carbon fixation and related
photosynthetic processes (Jacquemound et al..1996, 2000).
II.132
Figure 9. The radiance spectrum of a leaf or canopy expresses the presence and
abundance of the building blocks and products of photosynthesis. Elemental
and molecular contributions to the spectrum are labeled. This spectrum was
acquired over lowland Hawaiian rainforest using the Airborne Visible and
Infrared Imaging Spectrometer (AVIRIS).
II.133
Figure 10. Relative absorption intensity of chlorophyll-a and –b, as well as bulk
carotenoids, in Metrosideros polymorpha
What is the contribution of leaf N to reflectance signatures of tropical forests? Proteins have
absorptions spread throughout the shortwave-IR spectrum (Curran 1989) that are partially
obscured by water absorptions. In addition, chlorophyll is ~6% nitrogen by mass, and thus
chlorophyll and N tend to be broadly correlated with one another. Remote estimation of
chlorophyll tends to scale with N, with a range of correlation coefficients (r) from about 0.7
to 0.9 (Yoder and Pettigrew-Crosby 1995). Moreover, spectral analyses between leaf or
canopy spectra and nitrogen often show spectral correlations in the wavelength regions
associated with
both chlorophyll and protein-N (Martin and Aber 1997, Kokaly 2001, Smith et al.. 2003). But
are
correlations between spectra and N concentration direct, or are they indirect by way of leaf
chlorophyll and other leaf constituents? Leaf construction involves a stoichiometric balance
between chlorophyll, nitrogen, water and other biochemicals, all convolved with leaf
structure, especially SLA (Reich et al.. 1997, Evans and Poorter 2001). Leaf protein is
dominated by the enzyme Rubisco, which mediates carbon fixation in concert with light
capture by chlorophyll (Niinemets and Tenhunen 1997, Stylinski et al. 2000). Leaf water
concentrations scale with SLA, and SLA is correlated with leaf N per unit area (Wright et al.
2004). These leaf parameters are therefore not independent, biophysically or ecologically,
and thus the retrieval of leaf N tends to be wrapped up in the retrieval of the other parameters.
II.134
3.1 Spectral Transformations and parameter estimation
3.1.1 First derivative reflectance
A first difference transformation of the reflectance spectrum (FDR) calculates the slope
values from the reflectance and can be derived from the following equation (Dawson &
Curran, 1998). Derivative spectra have been widely used in spectroscopy and remote sensing.
Reflectance spectra may suffer from background signals and albedo effects. In contrast,
derivatives were shown to be less sensitive to variations of soil background, illumination, and
surface albedo (Demetriades-Shah et al.1990). The main disadvantage of derivatives is their
sensitivity to sensor noise. The decrease in the signal-to-noise ratio (SNR) with increasing
derivative order often limits the usage of higher order derivatives. The wavelength having
highest FDR is called red edge position (REP). If the value of REP shifts towards lower
wavelength, it is called blue shifting and indicates nutrient (nitrogen) stress and shifting
towards higher wavelength is called red shift and indicates healthy condition (Figure 11 a and
b).
(a)
(b)
Figure 11(a and b). Spectral Shift in healthy and stressed plant and (b) REP position for
low/nil and high N treated plot of wheat crop.
II.135
Figure 12. Estimation of canopy nitrogen, phosphorus, potassium, calcium and magnesium
concentration using four variables derived from continuum-removed absorption
features (From Mutanga et al.2003)
3.2 Parameter Retrieval from Hyper BRDF through radiative transfer modelling
Different methods to estimate canopy biophysical variables from reflectance data have been
developed and can be grouped into two approaches such as (1) statistical approach and (2)
physical process based approach (using Radiative Transfer Models). Using statistical
approach, many researchers have developed empirical relationships between vegetation
indices (VIs) and canopy biophysical variables. The equations defining such empirical and
semi-empirical relationships not only vary in the mathematical form (Linear, power,
Exponential, etc) but also in their empirical coefficients, depending upon the cultivars,
regions and the data normalizations approaches adopted. These methods are very simple but
the accuracy of biophysical variable estimation may be quite low. They are suffering from
severe limitations due to the lack of physics introduced in the retrieval technique and the
small amount of radiometric information they can exploit. Alternately, physical modeling
approach is based on the inversion of canopy reflectance models that describe the radiative
transfer in the canopy as a function of biophysical variables which characterize the canopy
architecture and the optical properties of vegetation elements and the soil. In the mid-80s, the
anisotropic properties were observed to be crucial for diagnosing plant canopy functioning.
Enhanced understanding of the physical processes that govern the interactions between light
and the canopy elements, bi-directional canopy reflectance (CR) models emerged for its
II.136
inversion issues on multidirectional data in the early 90s for retrieval of biophysical
parameters (Goel, 1987).
Inversion of bidirectional canopy reflectance (CR) models emerged as a promising alternative
for retrieval issues (Goel 1989; Myneni et al. 1991; Liang and Strahlar, 1993; Tripathi et al.
2006, 2009). The space borne instruments like POLDER, ADEOS, MISR, TERRA, etc were
designed to study both the spectral and directional characteristics of the earth surfaces. This
trend depicts one of the scientific stakes to come in remote sensing, which is to take
advantage of both the spectral and the directional signatures of vegetation in order to retrieve
the biophysical parameters.
Tripathi et al. (2006) conducted one field experiment with the objectives set as (i) to relate
canopy biophysical parameters and geometry to its bidirectional reflectance, (ii) to evaluate a
canopy reflectance model to best represent the radiative transfer within the canopy for its
inversion and (iii) to retrieve crop biophysical parameters through inversion of the model.
Two varieties of the mustard crop (Brassica juncea L) were grown with two nitrogen
treatments to generate a wide range of Leaf Area Index (LAI) and biomass. The reflectance
data obtained at 5nm interval for a range of 400-1100nm were integrated to IRS LISS –II
sensor’s four band values using Newton Cotes Integration technique. Biophysical parameters
were estimated synchronizing with the bi-directional reflectance measurements. The radiative
transfer model PROSAIL was used for its evaluation and to retrieve biophysical parameters
mainly LAI and Average Leaf Angle (ALA) through its inversion. Look Up Table (LUT) of
BRDF was prepared simulating through PROSAIL model varying only LAI (0.2 interval
from 1.2 to 5.4 ) and ALA (5° interval from 40 to 55°) parameters and inversion was done
using a merit function and numerical optimization technique given by Press et al. 1986. The
derived LAI and ALA values from inversion were well matched with observed one with
RMSE 0.521 and 5.57, respectively.
The radiative transfer model PROSAIL was used for retrieval of LAI, Chlorophyll (Cab) and
equivalent water thickness (Cw) of wheat crop of Trans Gangetic Plains through its inversion
(Fig 13). The model was calibrated for major parameters such as LAI, Cab, Cw and biomass
(Cm) and sensitivity analysis was performed. Inversion of PROSAIL model was carried out
for LAI, Cab and Cw using Look Up Table (LUT) approach. The merit function was
computed and used to best fit the measured data with the simulated one. Results revealed that
LAI, Cab and Cw, were very well retrieved with RMSE 0.3892, 4.307 and 0.0063
respectively when compared with measured values. The retrieved products were evaluated
with its corresponding regressed products through different vegetation indices. RMSE
between these regressed estimation and measured parameter values were 0.553, 5.204 and
0.01 for LAI, Cab and Cw respectively.
II.137
4.
5.
6.
7.
8.
9. EWT Chl
10.
11.
12.
Figure 13. The LAI, Equivalent Water Thickness and Total Chlorophyll content of wheat
crop in Trans Gangetic Plain of India retrieved from MODIS data through
radiative transfer modelling.
References
1. American Society of Photogrammetry, (1968). Manual of Color Aerial Photigraphy,
Falls Church, Va.
2. American Society of Photogrammetry, (1980). Manual of Photogrammetry, 4th ed.,
Falls Church, Va.
3. American Society of Photogrammetry, (1983). Manual of Remote Sensing, Vol. I, R. N.
Colwell (Ed.), Falls Church, Va.
4. Barbe, D.F. (1980). Charge – Coupled Devices. Topics in Applied Physics, Vol.38.
Springer- Verlag, Berlin.
5. Baret, F. and Guyot, G. (1991). Potential and limits of vegetation indices for LAI and
APAR assessment. Remote Sensing of Environment, 35, 161-173.
6. Barman, D., Sehgal, V.K., Sahoo, R.N. and Nagarajan, S. (2010). Relationship of
Bidirectional Reflectance of Wheat with Biophysical Parameters and its Radiative
Transfer Modeling Using PROSAIL, J. Ind. Soc. Remote Sens., 38, 141- 151.
7. Boegh, E., Soegaard, H., Broge, N., Hasager, C.B., Jensen, N.O. and Schelde, K.
(2002). Airborne multispectral data for quantifying leaf area index, nitrogen
concentration, and photosynthetic efficiency in agriculture. Remote Sensing of
Environment. 81,179−193.
8. Bonham-Carter, G.F. (1988). Numerical procedures and computer program for fitting an
inverted Gaussian model to vegetation reflectance data. Computer & Geosciences,
14,339-356.
9. Broge, N.H. and Leblanc, E. (2000). Comparing prediction power and stability of
broadband and hyperspectral vegetation indices for estimation of green leaf area index
and canopy chlorophyll density. Remote Sensing of Environment. 76, 156-172.
10. Carter, G.A.(1994). Ratios of leaf reflectances in narrow wavebands as indicators of
plant stress. International Journal of Remote Sensing, 15, 697−704.
II.138
11. Casa,R. and Jones, H.G. (2004). Retrieval of crop canopy properties: a comparison
between model inversion from hyperspectral data and image classification.
International Journal of Remote Sensing, 23, 1119-1130.
12. Cheng, Cheng Ji, (Ed.). (1989). Manual for the Optical-Chemical Processing of
Remotely Sensed Imagery. ESCAP / UNDP Regional Remote Sensing Programme
(RAS/86/141), Beijing, China.
13. Cho & Skidmore (2006). A new technique for extracting the red edge position from
hyperspectral data: The linea.r extrapolation method. Remote Sensing of Environment
101, 181–193.
14. Clark, R.N. and Roush, T.L. (1984). Reflectance spectroscopy: Quantitative analysis
techniquesfor remote sensing applications. Journal of Geophysical Research, 89 (B7),
6329-6340.
15. Clark, R.N. (1999). Spectroscopy of rocks and minerals and principles of spectroscopy.
In:Rencz, A. N. (Ed.): Remote sensing for the Earth sciences: Manual of remote
sensing, 3rd ed., Vol. 3 New York, Chichester, Weinheim. John Wiley & Sons, Inc.: 3-
58.
16. Curran, P.J.(1989). Remote sensing of foliar chemistry. Remote Sensing of
Environment, 30, 271-278.
17. Curran, P. J. (1985). Principles of Remote Sensing, Longman, London.
18. Dawson, T.P. and Curran, P.J. (1998). A new technique for interpolating the reflectance
red edge position. International Journal of Remote Sensing, 19(11) ,2133-2139.
19. Demetriades Shah, T.H., Steven, M.D. and Clark, J.A. (1990). High resolution
derivative spectra in remote sensing. Remote Sensing of Environment, 33, 55-64.
20. Despan, D. and Jacquemoud, S. (2004). Optical properties of soil and leaf: Necessity
and problems of modeling. In: Schönermark, M. von, Geiger, B., Röser, H. P. (Eds.):
Reflection properties of vegetation and soil. Berlin. Wissenschaft und Technik Verlag:
39-70.
21. Duckworth, J. (1998). Spectroscopic quantitative analysis. In: Workman, J.,
Springsteen, A. W. (Eds.): Applied spectroscopy. A compact reference for practitioners.
San Diego, New York, London. Academic Press: 93-164.
22. Elachi, Charles, (1987). Introduction to the Physics and Techniques of Remote Sensing,
John Wiley & Sons, New York.
23. Franklin, E. (Eds.): Remote sensing of forest environments - Concepts and case studies.
Boston,Dordrecht, London. Kluwer Academic Publishers: 279-300.
24. Franklin, J., Phinn, R., Woodcock, C.E. and Rogan, J. (2003). Rationale and conceptual
framework for classification approaches to assess forest resources and properties. In:
Wulder, M. A.,
25. Goel, N. (1989). Inversion of canopy reflectance models for estimation of biophysical
parameters from reflectance data, In Theory and Applications of Optical Remote
Sensing (G. Asrar, Ed), Wiley Interscience, New York.205-250.
26. Goel, N.S. (1987). Models of vegetation canopy reflectance and their use in estimation
of biophysical parameters from reflectance data, Remote Sensing Reviews, 1, 221.
II.139
27. Goel, N.S. (1989). Inversion of canopy reflectance models for estimation of biophysical
parameters from reflectance data, In Theory and Applications of Optical Remote
Sensing (G. Asrar, Ed), Wiley Interscience, New York, 205-250.
28. Guyot, G. (1990). Optical properties of vegetation canopies. In: Steven, M. D., Clark, J.
A. (Eds.): Applications of remote sensing in agriculture. London. Butterworths: 19-43.
29. Horler, D.N., Dockray, M. and Barber, J. (1983). The red edge of plant leaf reflectance.
International Journal of Remote Sensing. 4, 273-288.
30. Jacquemoud, S. and Baret, F. (1990). 'PROSPECT: A model of leaf optical properties
spectra, Remote Sensing of Environment. 34, 251-256.
31. Jacquemoud, S. (1993). Inversion of the PROSPECT + SAIL canopy reflectance model
from AVIRIS equivalent spectra: Theoretical study, Remote Sensing of Environment,
44(2-3), 281-292.
32. Jensen, J.R. (1996). Image Preprocessing: Radiometric and Geometric Correction. In :
Introductory Digital Image Processing. Prentice Hall publisher (Second Edition). 107-
135.
33. Kadupitiya, H.K., Sahoo, R.N., Gupta, V.K., Ahmed, N. and Ray, S.S. ( 2009).
Modeling soil chemical composition using hyperspectral reflectance data., In ISRS
Annual Convention and National Symposium on Advances in Geo-spatial Technologies
with special emphasis on Sustainable Rainfed Agriculture by Indian Soc. Of Remote
Sensing, Nagpur, Sept 17-19.
34. Kadupitiya, H.K., Sahoo, R.N., Ray, S.S., Chakraborty, D. and Ahmed, N. (2010).
Quantitative Assessment of Soil Chemical Properties using Visible (VIS) and Near-
infrared (NIR) Proximal. Hyperspectral data Tropical Agriculturist, 158, 41-60.
35. Kokaly, R.F. and Clark, R.N. (1999). Spectroscopic determination of leaf biochemistry
usingband-depth analysis of absorption features and stepwise multiple linear regression.
Remote Sensing of Environment, 67, 267-287.
36. Kruse, F.A., Lefkoff, A.B., Boardman, J.B., Heidebrecht, K.B., Shapiro, A.T., Barloon,
P.J. and Goetz, A.F.H. (1993). The Spectral Image Processing System (SIPS) -
Interactive Visualization and Analysis of Imaging spectrometer Data. Remote Sensing of
the Environment. 44, 145 – 163.
37. Kruse, F.A., Lefkoff, A.B. and Boardman, J.W. (1993). The Spectral Image Processing
System (SIPS) - Interactive visualization and analysis of imaging spectrometer data.
Remote Sensing of Environment. 44, 145-163.
38. Liang, S. and Strahler, A.H. (1993). Calculation of the angular radiance distribution for
a coupled atmosphere and canopy. IEEE Transactions on Geoscience and Remote
Sensing, 31(2), 491-502.
39. Lillesand, Thomas, M. and Kiefer, R.W. (1987). Remote Sensing and Image
Interpretation, 2nd ed., John Wiley & Sons, New York.
40. Lucas, N.S. and Curran, P.J. (1999). Forest ecosystem simulation modeling: The role of
remote sensing. Progress in Physical Geography. 23(3), 391-423.
41. Myneni,R.B. and Ross, J. (1991). Photon-Vegetation interactions: applications in
optical remote sensing and plant ecology, Springer-Verlag, New York. 565 .
42. Nilson, T. and Kuusk, A. (1989). A reflectance model for the homogeneous plant
canopy and its inversion. Remote Sensing of Environment. 27, 157-167.
II.140
43. Press, W.H., Flannery, B.P., Teukolsky, S.A. and Vetterling, W.T. (1986). Numerical
Recipes. Cambridge University Press, Cambridge. 994.
44. Ranjan, R., Sahoo, R.N., Tomar, R.K,, Nagarajan, S., Gupta, V.K. and Sharma, A.R.
(2008). Characterization of different crop residues and their differential affect on wheat
crop growth with hyper-spectral remote sensing. National Symposium on “Advances in
remote sensing technology with special emphasis on microwave remote sensing” (ISRS
symposium 2008), Dec. 18-20, Nirma University Campus, Ahmedabad.
45. Sabins, Floyd, F., Jr., (1987). Remote Sensing – Principles and Interpretations, 2nd ed.,
W. H. Freeman and Company, New York.
46. Sahoo, R.N. (2008) Potential Application of Hyperspectral Remote Sensing in
Agriculture In National Seminar on Hyperspectral Remote Sensing and Spectral
Signature Database Management System “Hyperspec 2008” Feb 14-15, 2008,
Annamalai University, Chidambaram, Tamilnadu.
47. Sahoo, R.N., Bhavanarayana, M., Panda, B.C., Arika, C.N. and Kaur, R. (2005). Total
information content as an index of soil moisture J. Indian Soc. Remote Sensing. 33(1),
17-24.
48. Sahoo, R.N., Tomar, R.K., Pandey, S., Chandana, P., Gupta, R.K., Sharama, R.K.,
Singh, S., Sehgal, V.K., Chakraborty, D. and Kalra, N. (2006). Remote sensing based
growth monitoring of RCT adopted wheat in rice based cropping system in Trans-
Gangetic Plain. Presented in “2nd International Rice Congress-Science, Technology and
Trade for Peace and Prosperity” 26th International Rice Research Conference, Oct 9-
13, New Delhi: 495.
49. Salisbury, J., Milton, N.M. and Walsh, P.A. (1987). Significance of non-isotropic
scattering from vegetation for geobotanical remote sensing. International Journal of
Remote Sensing, 8, 997-1009.
50. Santra, P., Sahoo, R.N., Das, B.S. and Gupta, V.K. (2009). Estimation of Soil Hydraulic
Properties using Spectral Reflectance in Visible and Near–Infrared Region
Geoderma,152, 338-349.
51. Schmidt and Skidmore (2002). Spectral discrimination of vegetation types in a coastal
wetland.Sciences. In: Rencz, A. N. (Ed.): Remote sensing for the Earth sciences:
Manual of remote Sensing of Environment. 67, 267-287.
52. Singh, D.K., Sahoo, R.N., Ahmed, N., and Gupta, V.K. (2008). Quantitative
Assessment of Soil Parameters from Hyperspectral Remote Sensing Data, National
Symposium on “Advances in remote sensing technology with special emphasis on
microwave remote sensing” (ISRS symposium 2008), Dec. 18-20, Nirma University
Campus, Ahmedabad.
53. Singh, D.K., Ahmed, N., Sahoo, R.N. and Gupta, V.K. ( 2008). Characterization of
some soil orders in relation to hyper-specteral reflectance data, National Symposium on
“Advances in remote sensing technology with special emphasis on microwave remote
sensing” (ISRS symposium 2008), Dec. 18-20, Nirma University Campus, Ahmedabad.
54. Singh, D.K., Ahmed, N., Sahoo, R.N. and Gupta, V.K. (2009). Determination of Some
Soil Properties using Hyper-Spectral Reflectance Data, Fourth World Congress on
“Conservation Agriculture”, Feb 4-7, New Delhi, India.
55. Toselli, F. (Ed.), (1989). Applications of Remote Sensing to Agrometeorology, Kluwer
Academic Publishers, Dordrecht, The Netherlands.
II.141
56. Tripathi R., Sahoo, R.N., Sehgal, V.K., Gupta ,V.K., Tomar, R.K. and Chopra, U.K.
(2009). Estimation of biophysical parameters of wheat from MODIS reflectance
through Radiative transfer modeling” In ISRS Annual Convention and National
Symposium on Advances in Geo-spatial Technologies with special emphasis on
Sustainable Rainfed Agriculture by Indian Soc. Of Remote Sensing, Nagpur, Sept 17-
19, 2009.
57. Tripathi, R., Sahoo, R.N., Sehgal, V.K., Tomar, R.K., Chakraborty, D., and Nagarajan
S. (2012). Inversion of PROSAIL Model for Retrieval of Plant Biophysical Parameters.
J. Indian Soc. Remote Sens, 40(1), 19 – 28.
58. Tripathi, R., Sahoo, R.N., Sehgal, V.K., Tomar, R.K., Pandey, S., Chakraborty, D. and
Kalra, N. ( 2006). Retrieval of plant biophysical parameters through inversion of
PROSAIL model. Proc. of SPIE Vol. 6411, 64110F-1 0277-786X/06/$15. doi:
10.1117/12.697954.
59. Tucker, C.J. ( 1980). Remote sensing of leaf water content in the near infrared. Remote
Sensing of Environment, 10, 23– 32.
60. Ustin, S. L., Smith, M. O., Jacquemoud, S. (1999). Geobotany: Vegetation mapping for
Earth Sciences. In: Rencz, A.N. (Ed.): Remote sensing for the Earth sciences: Manual of
remote sensing, 3rd ed., Vol. 3. New York, Chichester, Weinheim. John Wiley &
Sons.189-248.
61. Verhoef, W. and Bunnik, N.J.J.(1981). Influence of crop geometry on multispectral
reflectance determined by the use of canopy reflectance models. Proc.Intl.Colloq.
Signatures Remotely Sens. Objects, Augnon, France, 273-290.
62. Vermote, E.F. and Vermeulen, A.(1999). Atmospheric correction algorithm: Spectral
reflectances (MOD09). Algorithm Theoretical Background Document available online
at http://modarch.gsfc.nasa.gov/data/ ATBD/atbd_mod08.pdf (accessed 23 Feb. 2008)
63. Zarco, Tejada., P.J., Rueda, C.A. And Ustin, S.L.(2003). Water content estimation in
vegetation with MODIS reflectance data and model inversion methods. Remote Sensing
of Environment, 85(1), 109-124.
II.142
INTRODUCTION TO GEOGRAPHIC
INFORMATION SYSTEM
Prachi Misra Sahoo
I.A.S.R.I., New Delhi -110012
1. Introduction
Geographic Information System (GIS) is a computer based information system used to
digitally represent and analyse the geographic features present on the Earth surface and the
events that takes place on it. The meaning to represent digitally is to convert analog into a
digital form. "Every object present on the Earth can be geo-referenced", is the fundamental
key of associating any database to GIS. Here, term 'database' is a collection of information
about things and their relationship to each other, and 'geo-referencing' refers to the location of
a layer or coverage in space defined by the co-ordinate referencing system. Evolution of GIS
has transformed and revolutionized the ways in which planners, engineers, managers etc.
conduct the database management and analysis.
2. Defining GIS
A GIS is a system of hardware, software, data, people, organizations, and institutional
arrangements for collecting, storing, analyzing and disseminating information about areas of
the earth. It is also defined as an information system designed to work with data referenced
by spatial / geographical coordinates. In other words, GIS is both a database system with
specific capabilities for spatially referenced data as well as a set of operations for working
with the data. A Geographic Information System is a computer based system which is used to
digitally reproduce and analyse the feature present on earth surface and the events that take
place on it. In the light of the fact that almost 70% of the data has geographical reference as
it's denominator, it becomes imperative to underline the importance of a system which can
represent the given data geographically. The three perspectives of GIS are:
2.1 GIS as a Toolbox
GIS as a toolbox: if so then what kind of tools? This is classification based on functional
tasks of GIS like
1. Tools for automating spatial data (data capture via digitizing, scanning, remote
sensing, satellite geo-position system)
2. For storing spatial data (data bases and data structures)
3. For spatial data management/retrieval
4. For analysis (overlay, buffering, proximity, network functions, spatial statistics)
5. For display of spatial data and analysis results
2.2 GIS as an Information System
As Definition of GIS indicates GIS as a specialized information system stresses "spatially
distributed features (points, lines, areas), activities (physical and human-invoked), and events
(time).
2.3 GIS as an approach to Geographic Information Science
1. research on GIS (algorithms, analytical methods, visualization tools, user interfaces,
human-computer-human interaction)
2. research with GIS: GIS as a tool used by many substantive disciplines in their own
ways (anthropology, archeology, forestry, geology, engineering, business and
management sciences)
3. History of GIS
Work on GIS began in 1950s, but first GIS software came only in late 1970s from the lab of
the ESRI. Canada was the pioneer in the development of GIS as a result of innovations dating
back to early 1960s. Much of the credit for the early development of GIS goes to Roger
Tomilson. The events which took place in the development of GIS in chronological order are:
1. Early 1950s: thematic map overlay (superimposition of maps drafted at the same
scale)
2. 1959: use of transparent blacked-out overlays to find suitable locations (overhead)
3. 1960s: early computer mapping packages SURFACE II, SYMAP; early spatial data
banks (CIA’s World Data Bank )
4. 1960s: first GIS (Canada Geographic Information System, Minnesota Land
Management System)
5. 1970s: advances in algorithms and data structures to handle spatial data, Harvard Lab
for Computer Graphics (first modern GIS software)
6. 1980s: introduction of PC, advances in hardware, mature mainframe GIS software
7. 1990s: desktop GIS, Internet-based GIS services, proliferation of GIS applications
8. 2000 and beyond??? omnipresent GIS, wireless, networked, every-day applications
everywhere.
4. Components of GIS
GIS constitutes of five key components: Hardware, Software, Data, People, Method
4.1 Hardware
It consists of the computer system on which the GIS software will run. The choice of
hardware system range from 300MHz Personal Computers to Super Computers having
capability in Tera FLOPS. The computer forms the backbone of the GIS hardware, which
gets it's input through the Scanner or a digitizer board. Scanner converts a picture into a
digital image for further processing. The output of scanner can be stored in many formats e.g.
TIFF, BMP, JPG etc. A digitizer board is flat board used for vectorisation of a given map
objects. Printers and plotters are the most common output devices for a GIS hardware setup.
4.2 Software
GIS software provides the functions and tools needed to store, analyze, and display
geographic information. GIS softwares in use are MapInfo, ARC GIS, AutoCAD Map, etc.
The software available can be said to be application specific. When the low cost GIS work is
to be carried out desktop MapInfo is the suitable option. It is easy to use and supports many
GIS feature. If the user intends to carry out extensive analysis on GIS, ARC/Info is the
preferred option. For the people using AutoCAD and willing to step into GIS, AutoCAD Map
is a good option. The software for a geographical information system may be split into five
functional groups as mentioned below:
1. Data input and verification
2. Data storage and database management
III.144
3. Data transformation
4. Data output and presentation
5. Interaction with the user
4.3 Data
Geographic data and related tabular data can be collected in-house or purchased from a
commercial data provider. The digital map forms the basic data input for GIS. Tabular data
related to the map objects can also be attached to the digital data. A GIS will integrate spatial
data with other data resources and can even use a DBMS, used by most organization to
maintain their data, to manage spatial data.
4.4 People
GIS uses range from technical specialists who design and maintain the system to those who
use it to help them perform their everyday work. The people who use GIS can be broadly
classified into two classes. The CAD/GIS operator, whose work is to vectorise the map
objects. The use of this vectorised data to perform query, analysis or any other work is the
responsibility of a GIS engineer/user.
4.5 Method
And above all a successful GIS operates according to a well-designed plan and business
rules, which are the models and operating practices unique to each organization. There are
various techniques used for map creation and further usage for any project. The map creation
can either be automated raster to vector creator or it can be manually vectorised using the
scanned images. The source of these digital maps can be either map prepared by any survey
agency or satellite imagery.
5. Data in GIS
GIS stores information about the world as a collection of thematic layers that can be used
together. A layer can be anything that contains similar features such as customers, buildings,
streets, lakes, or postal codes. This data contains either an explicit geographic reference, such
as a latitude and longitude coordinate, or an implicit reference such as an address, postal
code, census tract name, forest stand identifier, or road name. There are two components:
spatial data that show where the feature is; and attribute data that provide information about
the feature. These are linked by the software. The fact that there are both spatial and attribute
data allows the database to be exploited in more ways than a conventional database allows, as
GIS provides all the functionality of the DBMS and adds spatial functionality.
5.1 Spatial Data
Spatial data is spatially referenced data that act as a model of reality. Spatial data represent
the geographical location of features for example points, lines, area etc. Spatial data typically
include various kinds of maps, ground survey data and remotely sensed imagery and can be
represented by points, lines or polygons.
5.2 Attribute Data
Attribute data refers to various types of administrative records, census, field sample records
and collection of historical records. Attributes are either the qualitative characteristics of the
spatial data or are descriptive information about the geographical location. Attributes are
stored in the form of tables, where each column of the table describes one attribute and each
row of the table corresponds to a feature.
III.145
6. The Nature of Geographical Data
1. Geographical position (spatial location) of a spatial object is presented by 2-, 3-
or 4-dimensional coordinates in a geographical reference system (e.g. Latitude and
Longitude).
2. Attributes are descriptive information about specified spatial objects. They often
have no direct information about the spatial location but can be linked to spatial
objects they describe. Therefore, it is often to call attributes "non-spatial" or
"aspatial" information.
3. Spatial relationship specifies inter-relationship between spatial objects (e.g.
direction of object B in relation to object A, distance between object A and B,
whether object A is enclosed by object B, etc.).
4. Time records the time stamp of data acquisition, specifies life of the data, and
identifies the locational and attribute change of spatial objects.
7. Representation of Spatial Data
In GIS, two types of data models represent spatial data: Vector model and Raster model
7.1 Vector Models
In vector models, objects are created by connecting points with straight line (or arcs) and area
is defined by sets of lines. Information about points, lines and polygons is encoded and stored
as a collection of x, y coordinate. Location of a point feature such as tubewell can be
described by a single x, y coordinate. Linear feature such as river can be stored as a collection
of point coordinates. Polygon feature, such as river catchment can be stored as a closed loop
of coordinates. Vector models are very useful for describing discrete features like data
represented by an area. Vector model is not much useful for describing continuously varying
features such as soil type.
III.146
Spatial Data: Vector format
(x1,y1)
Point - a pair of x and y coordinates
vertex
Line - a sequence of points
Node
In raster model, study area is divided into a regular grid of cells in which each cell contain
single value. Continuous data type, such as elevation, vegetation etc. are represented using
raster model
Cell size
Number
of
rows
NODATA cell
(X,Y)
Number of Columns
III.147
8. Methods of Data Input in GIS
8.1 Conversion of Existing Data
After receiving existing spatial data from government or private sources, GIS users often
need to convert the data to a format compatible with the GIS package. Existing data cannot
be easily converted because of a large variety of GIS packages and data formats in use. The
choice of a conversion method basically depends upon the specificity of the data format.
Some data formats are proprietary and require special translators for conversion of data from
one GIS package to another. Such a conversion method can be described as direct translation.
On the other hand, some data formats are neutral or public. In that case a GIS package needs
to have translators that can work with the neutral format for importing and exporting data.
8.2 Creating New Data
You can create new data by using primary data sources, such as satellite or GPS (Global
Positioning System) data, or secondary data sources, such as paper maps. The process of
converting paper maps to digital data is called digitizing, and it can be done using digitizer or
scanner technology.
8.2.1 Digitizing
Digitizing using a digitizing tablet is also called manual digitizing. The digitizing tablet has a
built-in electronic mesh and can sense the position of the cursor and transmit it to the
computer. The units of measurement on digitized coverages are in inches. Digitizing begins
with a set of tics, which are used later for converting the coverage to real-world coordinates.
Two considerations for manual digitizing are: point versus stream mode, and resolution and
accuracy.
8.2.2 Scanning
A scanner converts a paper map to a scanned file, which contains raster data with values of 1
and 0. Pixels with the value 0 represent lines scanned from the paper map, and pixels with the
value 1 represent non-inked areas. The scanned file is then converted to a coverage through
tracing. GIS packages such as ARC/INFO have algorithms that enable users to perform semi-
automatic tracing or manual tracing.
8.2.3 On-screen Digitizing
On-screen digitizing is an alternative to manual digitizing and scanning for limited digitizing
work such as editing or updating an existing coverage. On-screen digitizing is manual
digitizing on the computer monitor using a data source such as a DOQ as the background.
This is an efficient method for digitizing, for example, new trails or roads that are not on an
existing coverage but are on a new DOQ. Likewise, the method can be used for editing a
vegetation coverage based on new information from a new DOQ that shows recent clear-cuts
or burned areas. Obviously, the major shortcoming of this method is the resolution of the
computer monitor, which is much coarser than a digitizing table or a scanner.
9. Basic GIS Functionality
GIS software combines computer mapping functionality that handles and displays spatial
data, with database management system functionality to handle attribute data. The basic
functionality of GIS is as follows:
1. Querying both spatially and through attribute
III.148
2. Manipulating the spatial component of the data: for example, through changing
projections, rubber sheeting to join adjacent layers of data together, and calculating
basic statistics such as areas and perimeters of polygons
3. Buffering where all locations lying within a set distance of a feature or set of features
are identified
4. Data integration, either informally by simply laying one layer over another, or
formally through a mathematical overlay operation
5. Areal interpolation
III.149
11. Questions GIS can answers
Till now GIS has been described in two ways (i) Through formal definitions, and (ii) Through
technology's ability to carry out spatial operations, linking data sets together. However there
is another way to describe GIS by listing the type of questions the technology can (or should
be able to) answer. Location, Condition, Trends, patterns, Modelling, Aspatial questions,
Spatial questions. There are five type of questions that a sophisticated GIS can answer:
11.1 Location What is at………….?
The first of these questions seeks to find out what exists at a particular location. A location
can be described in many ways, using, for example place name, post code, or geographic
reference such as longitude/latitude or x/y.
11.2 Condition Where is it………….?
The second question is the converse of the first and requires spatial data to answer. Instead of
identifying what exists at a given location, one may wish to find location(s) where certain
conditions are satisfied (e.g., an unforested section of at-least 2000 square meters in size,
within 100 meters of road, and with soils suitable for supporting buildings)
11.3 Trends What has changed since…………..?
The third question might involve both the first two and seeks to find the differences (e.g. in
land use or elevation) over time.
11.4 Patterns What spatial patterns exists…………..?
This question is more sophisticated. One might ask this question to determine whether
landslides are mostly occurring near streams. It might be just as important to know how many
anomalies there are that do not fit the pattern and where they are located.
11.5 Modelling What if……………..?
"What if…" questions are posed to determine what happens, for example, if a new road is
added to a network or if a toxic substance seeps into the local ground water supply.
Answering this type of question requires both geographic and other information (as well as
specific models). GIS permits spatial operation.
11.6 Aspatial Questions
"What's the average number of people working with GIS in each location?" is an aspatial
question - the answer to which does not require the stored value of latitude and longitude; nor
does it describe where the places are in relation with each other.
11.7 Spatial Questions
"How many people work with GIS in the major centres of Delhi" OR " Which centres lie
within 10 Kms. of each other?" OR "What is the shortest route passing through all these
centres". These are spatial questions that can only be answered using latitude and longitude
data and other information such as the radius of earth. Geographic Information Systems can
answer such questions.
12. GIS as an Integrating Technology
In the context of these innovations, geographic information systems have served an important
role as an integrating technology. Rather than being completely new, GIS have evolved by
linking a number of discrete technologies into a whole that is greater than the sum of its parts.
GIS have emerged as very powerful technologies because they allow geographers to integrate
III.150
their data and methods in ways that support traditional forms of geographical analysis, such
as map overlay analysis as well as new types of analysis and modeling that are beyond the
capability of manual methods. With GIS it is possible to map, model, query, and analyze
large quantities of data all held together within a single database.
The importance of GIS as an integrating technology is also evident in its pedigree. The
development of GIS has relied on innovations made in many different disciplines:
Geography, Cartography, Photogrammetry, Remote Sensing, Surveying, Geodesy, Civil
Engineering, Statistics, Computer Science, Operations Research, Artificial Intelligence,
Demography, and many other branches of the social sciences, natural sciences, and
engineering have all contributed. Indeed, some of the most interesting applications of GIS
technology discussed below draw upon this interdisciplinary character and heritage.
13. Application Areas
GIS is now used extensively in government, business, and research for a wide range of
applications including environmental resource analysis, landuse planning, locational analysis,
tax appraisal, utility and infrastructure planning, real estate analysis, marketing and
demographic analysis, habitat studies, and archaeological analysis.
One of the first major areas of application has been in natural resources management,
including management of
1. wildlife habitat
2. wild and scenic rivers
3. recreation resources
4. floodplains
5. wetlands
6. agricultural lands
7. aquifers
8. forests
One of the largest areas of application has been in facilities management. Uses for GIS in
this area includes
1. locating underground pipes and cables,
2. balancing loads in electrical networks,
3. planning facility maintenance,
4. tracking energy use.
Local, state, and federal governments have found GIS particularly useful in land
management. GIS has been commonly applied in areas like
1. zoning and subdivision planning,
2. land acquisition,
3. environmental impact policy,
4. water quality management,
5. maintenance of ownership.
More recent and innovative uses of GIS have used information based on street-networks.
GIS has been found to be particularly useful in
1. address matching,
2. location analysis or site selection,
3. development of evacuation plans.
III.151
The range of applications for GIS is growing as systems become more efficient, more
common, and less expensive.
References
1. Antenucci, J.C., Brown, K., Croswell, P. L., Kevany, M. J., and Archer, H.N. (1991).
"Introduction," "Evolution of the Technology," and "Applications." Chaps. 1-3 in
Geographic Information Systems: A Guide to the Technology. New York: Van
Nostrand Reinhold.
2. Burrough, P.A. (1986). Principles of Geographic Information Systems for Land
Resources Assessment. Oxford: Oxford University Press.
3. Buckley, D. J. The GIS Primer: An Introduction to Geographic Information System.
http://www.innovativegis.com/basis/primer/primer.html.
4. Geographic Information System: Primer, Geospatial Training and Analysis Cooperative
http://geology.isu.edu/geostac/Field_Exercise/gisprimer/Frameset.html.
5. Huxhold, E. (1991). Information in the Organization. An Introduction to Urban
Geographic Information Systems. New York: Oxford University Press.
6. Star and John. (1990). Geographic Information Systems: An Introduction. Englewood
Cliffs, NJ: Prentice- Hall.
7. Tobler, W.R. (1959). Automation and Cartography. Geographical Review 49, 526-534.
III.152
DATA MODELS IN GIS
Prachi Misra Sahoo
I.A.S.R.I., New Delhi -110012
1. Introduction
GIS depicts the real world through models involving geometry, attributes, relations, and data
quality. Here the realization of models is described, with the emphasis on geometric spatial
information, attributes and relations. A prerequisite for describing the real world by use of GIS is
that the different type of geographical information can be stored in the computer. All the
operations in the computer are based on the storage and handling of numbers. This is why the
data stored in the computers is known as digital data. In GIS there is need to store graphical
figures, images, numerical values and plain text. All these forms of data must be converted into
digital representation. In principle, only two different numerical symbols or signals can be stored
in a computer – 0 and 1. The numerical system based on 0 and 1 is known as the binary system.
In order to separate different numbers form each other, the stream of 0’s and 1’s is divided into
groups of 8 bits. Each group is known as a byte. Decimal figures are stored with the help of 4
bytes (32 bits), based on logarithmic notation, by separate storage of the number’s mantissa and
exponent. Text is stored digitally with the help of a code system called ASCII (American
Standard Code for Information Interchange). Each number between 0 and 127 corresponds to
one sign on the computer’s key board. For example, uppercase letters A through Z are
represented by numbers between 97 and 90, and lowercase a through z by numbers between 97
and 122. To handle special national signs, a variation of the system has been developed to
accommodate up to 255 different signs.
Geometric presentations are commonly called digital maps. Strictly speaking, a digital map
would be peculiar because it would comprise only numbers (digits). By their very nature, maps
are analog, whether they are drawn by hand or machine, or whether they appear on paper or are
displayed on a screen. Technical, GIS does not produce digital maps – it produces analog maps
from digital map data. Nonetheless, the term digital map is now so widely used that the
distinction is well understood. Spatial information is presented geometrically in two ways: as
vector data in the form of points, lines, and areas (polygons), or as raster data in the form of
uniform, systematically organized cells. The vector model and raster model are discussed in the
following sections.
2. Basic Data Models in GIS
The data model represents a set of guidelines to convert the real world (called entity) to the
digitally and logically represented spatial objects consisting of the attributes and geometry. The
attributes are managed by thematic or semantic structure while the geometry is represented by
geometric-topological structure. There are two major types of geometric data model; vector and
raster model as shown in Fig. 1.
Figure 1. Vector and Raster data models
III.154
headwaters but as an area along its lower reaches. In this case, the width of the stream
determines how it should be represented on a map.
2.1.2 Topology of Vector Data Model
Topology expresses explicitly the spatial relationships between geometric objects. The vector
data model in ARC/INFO supports three basic topological concepts:
1. Connectivity: Arcs connect to each other at nodes
2. Area definition: An area is defined by a series of connected arcs
3. Contiguity: Arcs have directions and left and right polygons
III.155
Figure 3. The data structure of a line data model
III.156
2.1.3 Advantages and Disadvantages of Vector Data Models
The advantages of the vector data model are:
1. Good representation of entity data models. Compact data structure.
2. Topology can be described explicitly – therefore good for network analysis.
3. Coordinate transformation and rubber sheeting is easy.
4. Accurate graphic representation at all scales.
5. Retrieval, updating and generalization of graphics and attributes are possible.
The disadvantages of the vector data model are:
1. Complex data structures
2. Combining several polygon networks by intersection and overlay is difficult and requires
considerable computer power.
3. Display and plotting may be time consuming and expensive, particularly for high-quality
drawing, colouring, and shading.
4. Spatial analysis within basic units such as polygons is impossible without extra data
because they are considered to be internally homogeneous.
5. Simulation modeling of process of spatial interaction over paths not defined by explicit
topology is more difficult than with raster structures because each spatial entity has a
different shape and form.
2.2 Raster Format
Raster model uses regularly spaced grid cells in specific sequence. An element of the grid cell is
called a pixel (picture cell). The conventional sequence is row by row from the left to the right
and then line by line from the top to bottom. Every location is given in two dimensional image
coordinates; pixel number and line number, which contains a single value of attributes.
2.2.1 Geometry of Raster Data
The geometry of raster data is given by point, line and area objects as follows (see Figure 5)
a. Point Objects: A point is given by point ID, coordinates (i, j) and the attributes
b. Line Objects: A line is given by line ID, series of coordinates forming the line, and the
attributes
c. Area Objects: An area segment is given by area ID, a group of coordinates forming the area
and the attributes. Area objects in raster model are typically given by "Run Length" that
rearranges the raster into the sequence of length (or number of pixels) of each class as shown in
Figure 5.
The topology of raster model is rather simple as compared with the vector model as shown in
Figure 5. The topology of line objects is given by a sequence of pixels forming the line
segments. The topology of an area object is usually given by "Run Length" structure which
includes Start line no., (start pixel no., number of pixels), second line no., (start pixel no.,
number of pixels).
III.157
Figure 5. Geometry and Topology of Raster Data
2.2.2 Topology of Raster Data
One of the weak points in raster model is the difficulty in network and spatial analysis as
compared with vector model. For example, though a line is easily identified as a group of pixels
III.158
which form the line, the sequence of connecting pixels as a chain would be a little difficult in
tracing. In case of polygons in raster model, each polygon is easily identified but the boundary
and the node (when at least more than three polygons intersect) should be traced or detected.
a. Flow Directions
A line with directions can be represented by four directions called as the Rook's move in the
chess game or eight directions called as the Queen’s move, as shown in Figure 6 (a), (b), (c).
Figure 6 (c) shows an example of flow directions in the Queen's move. Water flow, links of a
network, roads etc. can be represented by the flow directions (or called Freeman chain code).
b. Boundary
Boundary is defined as 2 x 2 pixel window that has two different classes as shown in Figure 7
(a). If a window is traced in the direction shown in Figure 7 (a), the boundary can be identified.
c. Node
A node in polygon model can be defined as a 2 x 2 window that has more than three different
classes as shown in Figure 7 (b). Figure 7 (c) and (d) show an example of identification of pixels
on boundary and node.
III.159
Figure 7. Identification of Boundary and Node
2.2.3 Advantages and Disadvantages of Raster Data Models
The advantages of the raster data model are:
1. Simple data structures.
2. Location-specific manipulation of attribute data is easy.
3. Many kinds of spatial analysis and filtering may be used.
4. Mathematical modeling is easy because all spatial entities have a simple, regular shape.
III.160
5. The technology is cheap.
6. Many forms of data are available.
The disadvantages of the raster data model are:
1. Large data volumes.
2. Using large grid cells to reduce data volumes reduces spatial resolution, result in loss of
information and an inability to recognize phenomenological defined structures.
3. Crude raster maps are inelegant though graphic elegance is becoming much less of a
problem today. Coordinate transformations are difficult and time consuming unless
special algorithms and hardware are used and even then may result in loss of information
or distortion of grid cell shape.
2.3 Quadtree Data Model
Traditionally, the raster model is based on dividing the real world into equal-sized rectangular
cells. However, in many cases, it can be more practical to use a model with varying cell size.
Larger cells (lower resolution) may be used to represent larger homogeneous areas, and smaller
cells (higher resolution) may be used for more finely detailed areas. This approach, known as the
quad-tree representation, is a refinement of the block code method. In representing a given
areas, the aggregate amount of data involved is proportional to the square of the resolution (into
cells). Because the quad-tree model is a very practical concept, it is preferable for the storage of
both small and large volumes of data.
The quad-tree paradigm divides a geographical area into square cells of sizes varying from
relatively large to that of the smallest cell of the raster. Usually, the squares are then quartered
into four smaller squares. The quartering may be continued to a suitable level until a square is
found to be so homogeneous that it no longer needs to be divided, and the data on it can be
stored as a unit. A larger square may therefore comprise several raster cells having the same
values. However, homogeneous areas that are not square or do not coincide with the pattern of
squares employed may be further divided into homogeneous squares. The structure of the quad-
tree resembles an inverted tree, whose leaves are pointers to the attributes of homogeneous
squares and whose branch forks are pointers to smaller squares – hence the name quad-tree
(Figure 8).
2.3.1 Advantages and Disadvantages of Quadtree Data Models
The advantages of the quad-tree model are:
1. Rapid data manipulation, because homogeneous areas are not divided into the smallest
cells used.
2. Rapid search, because larger homogeneous areas are located higher up in the point
structure
3. Compact storage, because homogeneous squares are stored as units.
4. Efficient storage structure for certain operations, including searching for neighboring
squares or for a square containing a specific point.
III.161
5. The disadvantages of the quad-tree model are:
6. Establishing the structure requires considerable processing time.
7. Protracted processing may prolong alterations and updating
8. Data entered must be relatively homogeneous
9. Complex data may require more storage capacity than ordinary raster storage.
III.163
Figure 9. Different types of DEMs
III.164
References
1. Bernhardsen, T. (2002) Geographic Information Systems: An Introduction. John Wiley &
Sons, Inc.
2. Burrough, P.A. and McDonnell R.A. (1986) Principles of Geographic Information
Systems. Oxford: Oxford University Press.
3. Buckley, D. J. The GIS Primer: An Introduction to Geographic Information System.
http://www.innovativegis.com/basis/primer/primer.html.
III.165
SPATIAL DATA ANALYSIS OF VECTOR AND RASTER DATA
Prachi Misra Sahoo
I.A.S.R.I., New Delhi -110012
1. Introduction
The basic GIS functionality includes: Querying, Integrating and Manipulating Spatial Data. The
core functionality of GIS is that it combines computer mapping functionality that handles and
displays spatial data, with database management system functionality to handle attribute data.
The basic tools that GIS provides to the user that are not available in other types of software
packages are its basic functionality, which are as follows:
A user wants to join the two input line layers together, but due to inaccuracy in one or both of the
layers the lines do not match exactly. As part of the joining process the lines on the right-hand
layer are systematically distorted to allow them to join to the corresponding line on the left. This
distortion is at its maximum at the end points of the line and reduces as we move away from the
join. The layers can then be seamlessly joined.
Most GIS software packages will also calculate basic statistics about their spatial features.
Typical examples include calculating the length of lines, the area and perimeter of polygons, and
the distances between points. There are many examples of why these basic measures can be
useful. These include measuring distances along a transport network, the use of areas to calculate
population densities, and calculating the distances between settlements.
4. Buffering, Thiessen Polygons and Dissolving
There are times when rather than simply being interested in the locations of a type of feature, a
user is interested in the locations within a set distance of a feature. Examples of this might
include wanting to know all areas within (or outside of) 1km of a hospital, or areas within 10km
of a railway line, or within 5km of an urban area. Where information of this type is required a
buffering operation is used. Buffering takes a point, line, or polygon layer as input and produces
a polygon layer as its output as is shown in Figure 2.
III.167
Figure 2. Buffers around points, lines and a polygon
5. Data Integration
Manipulating the spatial component of a single layer of data is useful, but the full potential of
GIS lies in its ability to integrate data from a variety of layers. At a basic level this merely
involves combining layers on-screen to compare patterns. This might be as simple as taking a
raster scan of a map and placing a vector layer over the top. The raster layer provides a spatial
context for the features in the vector layer. Another option is to lay one vector layer on top of
another, for example to compare the pattern of roads with the location of farms to see which
farms lies near the major roads. Field boundaries might be a third layer added to this. This
approach goes beyond basic mapping, as querying the underlying attribute database allows a
detailed understanding of a multi-faceted study area to be developed. In this way an integrated
understanding of the problem can be derived from many (possibly highly disparate) sources. In
this section various single and multiple layer operation are mentioned.
5.1 Overlay Operations for Vector data
In addition to simply combining layers, querying them and comparing them, layers can be
combined to produce new layers through geometric intersections. This is called overlay. Any of
the three types of vector data can be overlaid with any of the others. An overlay operation
combines not only the spatial data but also the attribute data. This has many potential uses.
Overlay is one of the most important function of GIS. These involve combining different feature
type (point, line, area) from different layers to produce a new map containing features and
attributes of user interest. Thus it of three types:
i. Point-in-polygon Overlay
ii. Line-in-polygon Overlay
iii. Polygon-in-polygon Overlay
III.168
5.1.1 Point-in-polygon Overlay
Overlays point coverages on an polygon coverage. It computes contained in relationship and the
resulting point coverage contains new attributes ( point features assumes the polygon attributes
they lie within). For e.g. combining wells (point coverage) and planning districts (polygon
coverage) will help in queries like how many wells are there in each district.
5.1.2 Line-in-polygon Overlay
Overlays line coverage onto a polygon coverage. The line features in the output coverage
assumes the attribute of the polygon they lie within. The lines are broken at each area object
boundary
5.1.3 Polygon-in-polygon Overlay
Polygon overlay is a spatial operation which overlays one polygon coverage onto another to
create a new polygon coverage. The spatial locations of each set of polygons and their polygon
attributes are joined to derive new data relationships. The output coverage of such overlay
operation is a polygon coverage. Polygon-in-polygon overlay is of three types: Union,
Intersection, Identity.
Union The UNION procedure is equivalent to the boolean operator "OR" in which two or more
data layers are overlaid to produce a combined coverage. Every polygon in the output coverage
carries the attribute information of both the input and union coverage. The input and the union
coverages must be a polygon coverage.
Intersect The INTERSECT procedure is equivalent to the boolean AND operation. When the
two coverages are over-layed, only portion of the input coverage that falls inside the intersect
coverage will remain in the output coverage. While the intersect coverage must be a polygon, the
input coverage can be a line, polygon or point coverage. If the input coverage is a line coverage
the output coverage will also be a line coverage.
Identity The IDENTITY overlay function everything located within the boundaries of the input
coverage is collected in the output coverage. The boundary of the output coverage is identical to
the input coverage. The identity procedures applies to point, line and polygon coverages. If the
input coverage is a point feature the output coverage is also a point coverage.
5.2 Connectivity Analysis
Connectivity analysis is to analyze the connectivity between points, lines and areas in terms of
distance, area, travel time, optimum path etc. Connectivity analysis consists of the following
analyses.
i. Proximity Analysis
ii. Neighbourhood analysis
iii. Network Analysis
In raster overlay the values of cells in the output layer is calculated from the results of a
mathematical operation on the input layers. In this example the two input layers have been
added. Other operations such as multiplication and subtraction can also be used. When two
III.170
layers are combined using an overlay operation, the resulting layer will be at best as accurate as
the less accurate layer. Raster based overlay operation tools are:
i. Arithmetic functions (+, - , * , /)
6.2 Classification
Based on the number of classes before and after the classification, three types of classifications
can be differentiated:
a) one to one (1:1): The number of classes before is the same as the number of classes after the
classification process: there are no changes in the geometry of the spatial
objects, they have been re-assigned.
b) many to one (M:1): The number of classes after the classification is smaller than the number of
classes before the process: generalization, aggregation, merging
c) one to many (1:M): The number of classes after the classification process is more than the
those before the classification: in vector format spatial objects are split in
different objects; in raster format e.g. unique ID’s are assigned to each
pixel in the output map
6.3 (re)Classification
(re)Classification involves the selection and presentation of a selected layer of data based on the
classes or values of a specific attribute. It involves looking at an attribute, or a series of
attributes, for a single data layer and classifying the data layer based on the range of values of the
attribute. For examples: Reclassify a soil map into a PH map, classify an elevation map into
classes with intervals of 50 m.
7. Conclusions
GIS software provides extensive functionality that allows a user to approach his or her dataset in
a way that combines the spatial and attribute components of their data. This functionality leads to
added value being extracted from an existing dataset. All datasets have limitations, and the extra
functionality provided by GIS software allows us to use the data in ways that their creator would
never have envisaged. As a result, it is important to consider the limitations of all layers when
manipulating them with the GIS. It is also important to consider the limitations of the techniques
used on the data, particularly those that integrate data together. As long as the results of spatial
operations are understood within these limitations, GIS software provides new functionality that
should allow new understanding to be derived from spatially referenced data. This shows the
usefulness of the combined spatial and attribute data model used by GIS. This allows data to be
queried and integrated in ways that no other approach can manage. The key advantage of this is
that it allows the complexity of the data to be handled without undesirable simplification of the
data.
III.171
References
1. Bernhardsen, T. (2002). Geographic Information Systems: An Introduction. John Wiley
and Sons, Inc.
2. Burrough, P.A. and McDonnell, R.A. (1986). Principles of Geographic Information
Systems. Oxford: Oxford University Press.
3. Davis, B. E. (2001). GIS: A Visual Approach. Onward Press.
III.172
INTRODUCTION TO GLOBAL POSITIONING SYSTEM
Anil Rai
I.A.S.R.I., New Delhi -110012
1. Introduction
The Global Positioning System (GPS) is a U.S. space-based navigation system that provides
reliable positioning, navigation, and timing services to the army as well as civilian users on a
continuous worldwide basis - freely available to all. For anyone with a GPS receiver, the
system will provide location and time. GPS provides accurate location and time information
for an unlimited number of people in all weather conditions, day and night, anywhere in the
world. The GPS is made up of three parts:
1. Satellites orbiting the earth
2. Control and monitoring stations on earth and
3. The GPS receivers owned by users.
GPS satellites broadcast signals from space that are picked up and identified by GPS
receivers. Each GPS receiver then provides three-dimensional location (latitude, longitude,
and altitude) plus the time. Individuals may purchase GPS handsets that are readily available
through commercial retailers. Equipped with these GPS receivers, users can accurately locate
where they are and easily navigate to where they want to go, whether walking, driving,
flying, or boating. GPS has become a mainstay of transportation systems worldwide,
providing navigation for aviation, ground, and maritime operations. Disaster relief and
emergency services depend upon GPS for location and timing capabilities in their life-saving
missions. Everyday activities such as banking, mobile phone operations, and even the control
of power grids, are facilitated by the accurate timing provided by GPS. Farmers, surveyors,
geologists and countless others perform their work more efficiently, safely, economically,
and accurately using the free and open GPS signals.
2. History and Development
The GPS system was developed as a worldwide satellite based system by the U.S.
Department of Defence (DOD) to simplify and improve military and civilian navigation and
positioning. The system grew Out of the "space race" with the Soviet Union during the
1950s. By the 1960s the Air Force has developed a system in which several satellites with
accurate clocks could assist in determining the position of a vehicle moving on land or in the
air. In 1973, the Navy and Air Force programs combined and formed the Navigation
Technology Program, which eventually became NAVSTAR (NAVigation System with Time
And Ranging). The Russians also developed a GPS system called GLONASS (GLObal
Navigation Satellite System). More recently the European Union approved funding to
develop a GPS system called Galileo. Development and testing of the system began
following the first GPS satellite launch in 1974. These satellites were built by Rockwell
Collins and launched by the Air Force. Testing continued into the 1980s when GPS satellites
were to be among payloads carried by NASA Space Shuttle flights. The GPS program
suffered a major setback when shuttle launches were suspended following the 1986
Challenger accident. Several years passed until modifications could be made to the Delta II
launch vehicle, enabling it to carry GPS satellites. The GPS system became fully operational
on 8 December 1993 when the full constellation of 24 satellites, 21 operational and three in
reserve, became available. The cost to the Air Force (1973 - 2002) to develop the GPS
satellites (not including military user equipment or launch costs) is approximately $6.3
billion. It costs about $750 million annually to operate and maintain the constellation,
including research and development, as well as procurement for and replacement of satellites.
3. Anatomy of the term: ‘‘Global Positioning System”
Global: Anywhere on Earth. Well, almost anywhere, but not (or not as well):
1. inside buildings
2. underground
3. in very severe precipitation
4. under heavy tree canopy
5. around strong radio transmission
6. near powerful radio transmitter antennas
or, anywhere else not having a direct view of a substantial portion of the sky. The radio
waves that GPS satellites transmit have very short lengths- about 20 cm. A wave of this
length is good for measuring because it follows very short paths, unlike its longer cousins
such as AM and FM band radio waves that may bend considerably. Unfortunately, short
waves also do not penetrate matter very well, so the transmitter and the receiver must not
have much solid matter between them, or the waves are blocked, as light waves are easily
blocked.
Positioning: where are you? How fast are you moving and in what direction? In what
direction should you go to get to some other specific location, and how long would it take at
your speed to get there?
System: a collection of components with links among them. Components and links have
characteristics [1].
4. Structure /Segments of GPS
The Global Positioning System is comprised of three segments: satellite constellation/space
segment, monitoring network/ ground control segment and user receiving equipment/user
segment (Figure 4). Formal GPS Joint Program Office (JPO) programmatic terms for these
components are space, operational control and user equipment segments, respectively.
1. The satellite constellation contains the satellites in orbit that provide the ranging
signals and data messages to the user equipment.
2. The operational control segment (OCS) tracks and maintains the satellites in space.
The OCS monitors satellite health and signal integrity and maintains the orbital
configuration of the satellites. Furthermore, the OCS updates the satellite clock
corrections and ephemeredes as well as numerous other parameters essential to
determining user position, velocity, and time (PVT).
3. Lastly, the user receiver equipment performs the navigation, timing or other related
functions (e.g. surveying).
4.1 Space Segment
The Space Segment consists of the constellation of spacecraft and the signals broadcast by
them which allow users to determine position, velocity and time. The basic functions of the
satellites are to:
III.174
1. Receive and store data transmitted by the Control Segment stations.
2. Maintain accurate time by means of several onboard atomic clocks.
3. Transmit information and signals to users on two L-band frequencies.
Several constellations of GPS satellites have been deployed, and several more are planned.
The experimental satellites, the so-called "Block I" satellites, were built by the Rockwell
Corporation. The first was launched in February 1978, and the last of the eleven satellite
series (one exploded on the launch pad) was launched in 1985. The operational series of GPS
satellites, the "Block II" and "Block IIA" satellites were also built by the Rockwell
Corporation. The 20 replacement "Block IIR" series of satellites, first launched in 1997, are
built by the General Electric Corporation. The "Block IIF" series are still in the design phase
and may, for example, incorporate an additional civilian transmission frequency. The
operational satellite I.D.s are separated into three space vehicle numbering series: SVN 13
through 21 for the Block II, SVN 22 through 40 for Block IIA, and SVN 41 and above for the
Block IIR satellites.
The Space Segment is an earth-orbiting constellation of 24 active and five spare GPS
satellites circling the earth in six orbital planes. Each satellite is oriented at an angle of 55
degrees to the equator. The nominal circular orbit is 20,200-kilometer (10,900 nautical miles)
altitude. Each satellite completes one earth orbit every twelve hours (two orbits every 24
hours). That's an orbital speed of about 4 km per second.
Each satellite has a design life of approximately 10 years, weighs about 845 kg, and is
approximately 17 feet across with its solar panels extended. Older satellites (designated
Block II/IIA) still functioning are equipped with 2 cesium, and 2 rubidium atomic clocks.
Newer satellites (Block IIR) are equipped with rubidium atomic clocks. All satellites also
contain 3 nickel-cadmium batteries for backup power when a satellite is in earth eclipse (out
of view of the sun).
Each GPS satellite (Figure 1) transmits a unique navigational signal centred on two L-band
frequencies of the electromagnetic spectrum, permitting the ionospheric propagation effect on
the signals to be eliminated. At these frequencies the signals are highly directional and so are
easily reflected or blocked by solid objects. Clouds are easily penetrated, but the signals may
be blocked by foliage (the extent of blockage is dependent on the type and density of the
leaves and branches). The satellite signal consists of the following components:
1. The two L-band carrier waves.
2. The ranging codes modulated on the carrier waves.
3. The so-called "navigation message".
III.175
Several different notations are used to refer to the satellites in their orbits. One particular
notation assigns a letter to each orbital plane (i.e., A, B, C, D, E, and F) with each satellite
within a plane assigned a number from 1 to >=4. Thus, a satellite referenced as B3 refers to
satellite number 3 in orbital plane B. A second notation used is a NAVSTAR satellite number
assigned by the U.S. Air Force. This notation is in the form of a space vehicle number (SVN)
11 to refer to NAVSTAR satellite 11.
4.2 Control Segment (CS)
The CS has responsibility for maintaining the satellites and their proper functioning. This
includes maintaining the satellites in their proper orbital positions (called station keeping) and
monitoring satellite health and status. The CS (Figure 2) also monitors the satellite solar
arrays, battery power levels also. There are five ground facility stations: Hawaii, Colorado
Springs, Ascension Island, Diego Garcia and Kwajalein. All are owned and operated by the
U.S. Department of Defence and perform the following functions:
1. All five stations are Monitor Stations, equipped with GPS receivers to track the
satellites. The resultant tracking data is sent to the Master Control Station.
2. Colorado Springs is the Master Control Station (MCS), where the tracking data are
processed in order to compute the satellite ephemerides and satellite clock corrections.
It is also the station that initiates all operations of the space segment, such as
spacecraft manoeuvring, signal encryption, satellite clock-keeping, etc.
3. Three of the stations (Ascension Is., Diego Garcia, and Kwajalein) are Upload
Stations allowing for the uplink of data to the satellites. The data includes the orbit
and clock correction information transmitted within the navigation message, as well
as command telemetry from the MCS.
Overall operation of the Control and Space Segments is the responsibility of the U.S. Air
Force Space Command, Second Space Wing, Satellite Control Squadron at the Falcon Air
Force Base, Colorado.
The GPS satellites travel at high velocity (4 km/sec), but within a more or less regular orbit
pattern. After a satellite has separated from its launch rocket and it begins orbiting the earth,
its orbit is defined by its initial position and velocity, and the various force fields acting on
the satellite. In the case of the gravitational field for a spherically symmetric body this
produces an elliptical orbit which is fixed in space – the Keplerian ellipse. Due to the effects
of the other, non-spherical gravitational components of the earth's gravity field, and non-
gravitational forces, which perturb the orbit, the actual trajectory of the satellite departs from
the ideal Keplerian ellipse.
The most significant forces that influence satellite motion are:
1. the spherical and non-spherical gravitational attraction of the earth,
2. the gravitational attractions of the sun, moon, and planets (the "third body" effects),
3. atmospheric drag effects,
4. solar radiation pressure (both direct and albedo effects).
III.176
Figure 2. Monitor Stations and Master Control Stations of GPS.
4.3 User Segment
The user receiving equipment, typically referred to as a GPS receiver (Figure 3), processes
the L-band signals transmitted from the satellites to determine PVT. There has been a
significant evolution in the technology of GPS receiving sets since they were initially
manufactured in the mid-70. Initially, they were large, bulky and heavy analog devices
primarily used for military purposes. With today’s technology, a GPS receiver of comparable
or more capability typically weighs a few pounds or ounces, and occupies a small volume.
The smallest of todays are those of a wrist watch size, while the largest is a naval shipboard
unit (weighing about 32 kgs). The basic structure of a receiver is the antenna, the receiver and
processor, the display and a regulated dc-power supply. These receivers can be mounted in
ships, planes and cars, and provide exact position information, regardless of weather
conditions.
III.177
5. GPS Satellite Constellation and Signals
5.1 Satellite Constellation
The operational Block II/IIA satellite constellation was to be fully deployed by the late
1980’s.However, a number of factors, the main one being the Space Shuttle Challenger
disaster (28 January 1986), has meant that the GPS system only became operational in the
1990’s as far as most users were concerned. Full Operational Capability was declared on 17
July 1995 – 24 Block II/IIA satellites operating satisfactorily. At an altitude of approximately
20,200km, a constellation of 24 functioning GPS satellites (Figure 5) is sufficient to ensure
that there will always be at least four satellites visible, at all unobstructed sites on the globe.
Typically there are 6 to 10 satellites visible most of the day. The U.S. Department of Defence
has undertaken to guarantee 24 satellite coverage 70% of the time, and 21 satellite coverage
98% of the time.
As the GPS satellites are in nearly circular orbits, at an altitude of approximately 20,200km
above the earth, this has a number of consequences:
1. Their orbital period is approximately 11hrs 58mins, so that each satellite makes two
revolutions in one sidereal day (the period taken for the earth to complete one rotation
about its axis with respect to the stars).
2. At the end of a sidereal day (23hrs 56mins in length) the satellites are again over the
same position on earth.
3. Reckoned in terms of a solar day (24hrs in length), the satellites are in the same
position in the sky about four minutes earlier each day.
III.178
Table 1: GPS signal components
III.179
A copy of the C/A code for a given satellite might look like this:
1 0 0 0 1 1 0 1 0 0 1 0 1 1 1 1 0 1 1 0 0 0 1…
and it is a total of 1,023 bits. Then the sequence starts again. The sequence above probably
looks random to us. It is in fact called a pseudorandom noise code- the term noise coming
from the idea that an aural version of it would greatly resemble static one might hear on a
radio. The acronym is PRN.
Now the question is “how does the receiver use the 0’s and 1,s to determine the range from
the satellite to the receiver?”
The PRN code is any thing but random. A given satellite uses a computer program to
generate its particular code. The GPS receiver essentially uses a copy of the same computer
program to generate the identical code. Further, the satellite and the receiver begin the
generation of the code at exactly the same moment in time.
The receiver can therefore determine its range from the satellite by comparing the two PRN
sequences (the one it receives and the one it generates). The receiver first determines how
much the satellite signal is delayed in time, and then, since it knows the speed of radio waves,
it can calculate how far apart the two antennas are in space.
As an example (using letters rather than bits so we can have a more obvious sequence, and
cooking the numbers to avoid explaining some unimportant complications), suppose the
satellite and the receiver each began, at 4:00 P.M., to generate one hundred letters per second:
G J K E T Y U O W V W T D H K...
The receiver would then look at its own copy of this sequence and the one it received from
the satellite. Obviously its own copy would start at 4:00, but the copy from the satellite
would come along after that,because of the time it took the signal to cover the distance
between the antennas.Below is a graphic illustration of what the two signals might look like
to the computer in the receiver:
Receiver: GJKETYUOWVWTDHK...
Satellite: GJKETYUOWVWTDHK...
The receiver would attempt to match the signals. We can see that the signal from the satellite
began to arrive seven letters later than 4:00; the receiver’s microcomputer could therefore
determine that it took 7/100 of a second for the signal from the satellite to reach the receiver
antenna. Since the radio wave travels at about 300,000 kilometers per second, the time
difference would imply that the satellite was 21,000 (that is,7/100*300,000) kilometers from
the antenna.[1]
5.2.2 Navigation Message
In addition to the C/A code navigational information is modulated into both the signals. The
information contains data like satellite orbits, clock corrections and other system parameters
(information about the status of the satellites). These data are constantly transmitted by each
satellite. From these data receiver gets its date, the approximate time and the position of the
III.180
satellites. The complete data signal consists of 37500 bit and at a transmission rate of 50 bit/s.
A total of 12.5 minutes is necessary to receive the complete signal. This time is required by a
GPS receiver until the first determination of a position is possible, if no information about the
satellites is stored or the information is outdated. The data signal is divided into 25 frames
(Figure 7), each having a length of 1500 bit (meaning an interval of 30 seconds for
transmission).
The 25 frames are divided into subframes (300 bit, 6 sec.), which are again divided into 10
words each (30 bit, 0.6 sec). The first word of each subframe is the TLM (telemetry word). It
contains a synchronization pattern, used by the receiver to help synchronise itself with the
navigation message and thus be able to correctly decode the data content. The next word is
the HOW (hand over word), which contains the number of counted z-epochs. This is the time
according to the satellite’s clock when the end of the subframe will be transmitted. The rest
of the first subframe contains data about status and accuracy of the transmitting satellite as
well as clock correction data. The second and third subframes contain ephemeris parameters.
Subframes 4 and 5 contain the so-called almanac data which include information about orbit
parameters of all satellites, their technical status and actual configuration, identification
number and so on. Subframe 4 contains data for the satellites number 25 – 32, ionospheric
correction data, special information and UTC time information; subframe 5 contains almanac
data for the satellites 1 – 24. The first three subframes are identical for all 25 frames. Every
30 seconds the most important data for the position determination are transmitted with these
three subframes. From the almanac data the GPS receiver identifies the satellites that are
likely to be received from the actual position. The receiver limits its search to these
previously defined satellites and hence this accelerates the position determination.
6. Working of GPS
The GPS system consists of three pieces. There are the satellites that transmit the position
information, there are the ground stations that are used to control the satellites and update the
information, and finally there is the receiver that we purchased. It is the receiver that collects
data from the satellites and computes its location anywhere in the world based on information
it gets from the satellites. There is a popular misconception that a GPS receiver somehow
sends information to the satellites but this is not true, it only receives data. So, just how is it
able to compute its position? Two views are there-
6.1 Geometric View
The basic concept of GPS positioning is that of positioning-by-ranges. The geometrical
principles of positioning can be demonstrated in terms of the intersection of loci. In the two-
dimensional case, a measured range to a known point constrains the position to lie on circle
with the measured range as radius. In three dimensions a measured range to a known point
III.181
constrains the position in 3-D space to lie on the surface of a sphere centred at the known
point (Figure 8), with radius being the measured distance. In the case of GPS, the distance
measurement is made to a satellite with known position (coordinates are obtained from the
satellite ephemeris data transmitted within the navigation message), however the principle
applies to any range measuring positioning system, terrestrial or satellite-based. In two
dimensions, position can be defined as the intersection of two circles, involving distances d1
and d2 to two known points, as shown below (Figure 9). Note that there are two possible
solutions, only one of which is correct. In general one solution can be discarded rather easily
as the point lies in space.
In the three-dimensional case, the intersection of three spheres describes two points in space,
only one of which is correct (Figure 10). Hence, a minimum of three ranges are required, to
three separated known points, in order to solve the 3-D position problem. The quality of the
positioning solution is dependent, amongst other things, on the accuracy with which the
ranges can be measured and the geometry of the intersection.
III.182
Figure 10. Intersection of Surfaces of Position Based on Range Measurements.
If the point being positioned is stationary, the two (or three) ranges do not need to be
measured simultaneously. If the point is moving however, all ranges must be measured
simultaneously (or over an interval of time during which the point has not moved by an
amount greater than the uncertainty of the "fix"). Because the GPS constellation was
designed to ensure at least four satellites are always visible anywhere on the earth, satellite
positioning using simultaneously measured ranges is the basic positioning strategy for most
navigation applications. However, there is still the issue of how to account for measurement
biases, as the technology used for making GPS range measurements does not give calibrated
distance from the receiver to the satellite. Disturbing influences and errors in fact contaminate
the range measurements to an unacceptable degree, and hence the basic positioning principle
is modified in several ways to satisfy the varying levels of accuracies required by different
applications.
6.2 Mathematical View
The observation equation for a receiver-clock-biased range is:
where c is the velocity of electromagnetic radiation in a vacuum (or simply the "velocity of
light"), erc is the receiver clock error (assume satellite clock time is "true" time) at time of
reception tr, P is the measured range and p is the true "geometric" range. Each observation
made by the receiver can be parameterised as follows:
where xs, ys, zs is the coordinate of the satellite and x, y, z is the coordinate of the receiver.
As the 3-D coordinate of the satellite is known, then each measurement P contains four
parameters which may be considered unknown: the 3-D coordinate of the receiver (x, y, z)
and the receiver clock error (erc). By making four measurements, to four different satellites,
the following system of equations is obtained:
III.183
(xs1 – x)2 + (ys1 – y)2 + (zs1 – z)2 = (P1 – erc.c)2
(xs2 – x)2 + (ys2 – y)2 + (zs2 – z)2 = (P2 – erc.c)2
(xs3 – x)2 + (ys3 – y)2 + (zs3 – z)2 = (P3 – erc.c)2
(xs4 – x)2 + (ys4 – y)2 + (zs4 – z)2 = (P4 – erc.c)2 ... (3)
Equation (3) has a unique solution. So from the solution of these equations we can easily
determine the coordinate of the receiver as well as the correct time (UTC time).
If more than four measurements are made the method of Least Squares, or other estimation
techniques, can be used to determine the optimum solution. Does the receiver clock error
have to be estimated at each epoch? That depends upon such factors as:
1. How well the clock error is estimated.
2. How often the position solution is carried out.
3. The stability of the clock.
6.2.1 Satellite Clock Bias
As the satellite clock error is the largest source of GPS measurement bias it deserves closer
study. Under the assumption that the satellite clock error is an unknown quantity, the
observation equation for such a satellite-biased range:
esc is the satellite clock error caused by the satellite atomic clock not being synchronized to
"true" time (GPST/UTC). P is the measured range, p is the true range and Tr is the time of
transmission. Each observation made by the receiver can be parameterised as in equation (2),
except for the replacement of the term esc for erc:
The 3-D coordinate of the satellite signal transmitter ( xs, ys, zs ) is known, hence in the case
of three range observations there are six unknowns in the system: the 3-D coordinate of the
receiver ( xr1, yr1, zr1 ) and the three satellite clock error terms ( esc1, esc2, esc3 ). Hence, at first
glance, six satellite-biased range observations are required to solve this positioning problem.
It is not, however, possible simply to make observations to more satellites as each new
observation introduces a new satellite clock parameter. There are two options for overcoming
this dilemma.
It is possible to take advantage of the fact that all observations made to a particular satellite
are biased by the same amount (if made at the same time, or close enough together so that the
satellite clock error can be assumed to have not changed by a significant amount). If three
range observations are made from another station, whose coordinate is known ( xr2, yr2, zr2 ),
then it is possible to obtain a system of six equations in six unknowns:
III.184
For which a unique solution can be obtained. In a conceptual sense this is the basis of
differential GPS positioning,
The other strategy for accounting for satellite clock error is for the GPS operators to
periodically determine the clock error. As the satellite clocks have significantly better long-
term drift characteristics than the receiver clocks, a suitable clock error model could be a time
polynomial:
III.185
7.2 Non-correctable Errors
Non-correctable errors cannot be correlated between two GPS receivers that are located in the
same general area. Sources of non-correctable errors include receiver noise, which is
unavoidably inherent in any receiver, and multipath errors.
7.2.1 Multipath Error
This occurs when the GPS signal is reflected off objects such as tall buildings or large rock
surfaces before it reaches the receiver. This increases the travel time of the signal, thereby
causing errors.
The error sources and the approximate error range are given below:
III.186
same GPS radio signals its receiving. Here the GPS receiver triangulates its position with
GPS as x+30, and y+60. The DGPS receiver provides the correction factor to the other GPS
receiver’s processor, which calculates its correct position by subtracting 5 from ‘x’ co-
ordinate and adding 3 with the ‘y’ co-ordinate. The user can generally get accurate position
fixes within a few meters or less using DGPS.
Many high-end GPS receivers have built in DGPS capability, while some low-end receivers
(including some Garmin models) can be configured for DGPS using add on hardware. There
are a number of free and subscription services available to provide DGPS corrections. The
U.S. and Canadian Coast Guards and U.S. Army Corps of Engineers transmit DGPS
corrections through marine beacon stations.
The Wide Area Augmentation System (WAAS) is being developed by the Federal Aviation
Administration (FAA) as another kind of highly advanced DGPS.
9. Applications of GPS
9.1 Timing
In addition to longitude, latitude, and altitude, the Global Positioning System (GPS) provides
a critical fourth dimension – time. Each GPS satellite contains multiple atomic clocks that
contribute very precise time data to the GPS signals. GPS receivers decode these signals,
effectively synchronizing each receiver to the atomic clocks. This enables users to determine
the time to within 100 billionths of a second, without the cost of owning and operating atomic
clocks. Precise time is crucial to a variety of economic activities around the world.
Communication systems, electrical power grids, and financial networks all rely on precision
timing for synchronization and operational efficiency. The free availability of GPS time has
enabled cost savings for companies that depend on precise time and has led to significant
advances in capability.
9.2 Roads and Highways
It is estimated that delays from congestion on highways, streets, and transit systems
throughout the world result in productivity losses in the hundreds of billions of dollars
annually. Other negative effects of congestion include property damage, personal injuries,
increased air pollution, and inefficient fuel consumption. The availability and accuracy of the
Global Positioning System (GPS) offers increased efficiencies and safety for vehicles using
highways, streets, and mass transit systems. Many of the problems associated with the routing
and dispatch of commercial vehicles is significantly reduced or eliminated with the help of
GPS. Many nations use GPS to help survey their road and highway networks, by identifying
the location of features on, near, or adjacent to the road networks. These include service
stations, maintenance and emergency services and supplies, entry and exit ramps, damage to
the road system, etc. The information serves as an input to the GIS data gathering process.
This database of knowledge helps transportation agencies to reduce maintenance and service
costs and enhances the safety of drivers using the roads.
9.3 Space
The Global Positioning System (GPS) is revolutionizing and revitalizing the way nations
operate in space, from guidance systems for crewed vehicles to the management, tracking,
and control of communication satellite constellations, to monitoring the Earth from space.
Benefits of using GPS include:
Navigation Solutions -- providing high precision orbit determination, and minimum ground
control crews, with existing space-qualified GPS units.
III.187
Attitude Solutions -- replacing high cost on-board attitude sensors with low-cost multiple
GPS antennae and specialized algorithms.
Timing Solutions -- replacing expensive spacecraft atomic clocks with low-cost, precise time
GPS receivers.
9.4 Aviation
Aviators throughout the world use the Global Positioning System (GPS) to increase the safety
and efficiency of flight. With its accurate, continuous, and global capabilities, GPS offers
seamless satellite navigation services that satisfy many of the requirements for aviation users.
Space-based position and navigation enables three-dimensional position determination for all
phases of flight from departure, en route, and arrival, to airport surface navigation. New and
more efficient air routes made possible by GPS are continuing to expand. Vast savings in
time and money are being realized. In many cases, aircraft flying over data-sparse areas such
as oceans have been able to safely reduce their separation between one another, allowing
more aircraft to fly more favourable and efficient routes, saving time, fuel, and increasing
cargo revenue.
9.5 Agriculture
The development and implementation of precision agriculture or site-specific farming has
been made possible by combining the Global Positioning System (GPS) and geographic
information systems (GIS). These technologies enable the coupling of real-time data
collection with accurate position information, leading to the efficient manipulation and
analysis of large amounts of geospatial data. GPS-based applications in precision farming are
being used for farm planning, field mapping, soil sampling, tractor guidance, crop scouting,
variable rate applications, and yield mapping. GPS allows farmers to work during low
visibility field conditions such as rain, dust, fog, and darkness.
9.6 Surveying and Mapping
Using the near pinpoint accuracy provided by the Global Positioning System (GPS) with
ground augmentations, highly accurate surveying and mapping results can be rapidly
obtained, thereby significantly reducing the amount of equipment and labour hours that are
normally required of other conventional surveying and mapping techniques. Today it is
possible for a single surveyor to accomplish in one day what used to take weeks with an
entire team. GPS is unaffected by rain, wind, or reduced sunlight, and is rapidly being
adopted by professional surveyors and mapping personnel throughout the world.
10. GPS and Precision Agriculture
Precision Agriculture is doing the right thing, at the right place, at the right time. Knowing
the right thing to do may involve all kinds of high tech equipments and fancy statistics or
other analysis.
In this context, GPS becomes part of precision agriculture. For analysis and processing of
remote Sensed images requires ground truth information, collected in the field, at a variety of
sites and often at various times throughout the crop production season. Conventionally this
data has been manually recorded on field sheets, air photos or paper maps and considerable
time and effort is required to convert it to digital format for use in remote sensing or GIS. For
image analysis the ground data must be digitized in order to create a mask for training the
software to recognize different conditions and classify the remote sensing imagery. Now we
have an interactive, portable system to record field data directly into a digital database
consisting of yield, soil, road, water and contour maps overlain on air photos or remote
sensing imagery. A GPS receiver is linked to a note book computer displaying appropriate,
III.188
pre loaded information layers, and a software package then combines incoming GPS signals
with the displayed data to allow the user to see where they are with respect to the map
components.
In precision agriculture, our main objective is to get more and more output by providing
optimum input. In agricultural Production soil is the media on which a seed is shown or a
plant is planted. So soil is very important from production point of view.
10.1 Soil Sampling by using GPS
Soil Sampling is like the foundation of a house. No matter how much effort we put into
building the house, the house is only as good as the foundation. The same principle applies to
precision agriculture. Whether growing forage, feed, food, or fibre, plant growth depends on
soil conditions and soil quality. To effectively manage soil-plant interrelationships, soils
information is very important.
There are two basic types of grid sampling (Figure 11) used to collect soils data for precision
agriculture
1. Area sampling (grid cell)
2. Point sampling with interpolation (grid point)
Grid sampling is used for precision agriculture because it is simple and does not require soil
science mapping experience. Once the soil data has been collected, the data can be displayed
and analyzed.
Determining and mapping the variations in soil characteristics across a field requires and
accurate knowledge of the position that the samples were taken. Whichever grid sampling
method is used the coordinate location of the soil sample should be accurate for developing a
soils data layer and for navigating back to those locations for re-sampling.
This requires the use of a GPS receiver and a source of differential corrections so the
producer can acquire an accurate (1-2 meter) horizontal position that represents the soil
sample location. Having acquired the coordinates, the position can be entered into a database
while in the field.
After the soil sample location has been accurately acquired and entered into a database, all
the physical soils data (texture, pH, nutrients, etc.) can be tagged to the coordinate location.
III.189
11. A Case Study: “Use of Remote Sensing Technology in Crop Yield Estimation
Surveys-2”.
11.1 Objective of the Study
a) To test the methodology of stratification based on satellite data in the form of vegetation
indices in crop yield surveys.
b) To obtain improved estimator of crop yield from crop yield estimation surveys using post
stratification based on satellite spectral data.
11.2 Study Area
The study was conducted for Rohtak district of Haryana state which is one of the major wheat
growing areas having acreage of more than 66% under wheat crop. Geographically the
district lies between 76015` to 77000` east longitude and 28020` to 29005` north latitude. The
total area of the district is 3911 sq. Kms. The land of the district is generally plane. Three
railway lines, one national highway, Western Yamuna main canal and Bohar and Bhalaut
branches of Western Yamuna main canal are present in the district. All these features provide
a great aid in identification of villages selected for crop cutting experiment and interpretation
of satellite spectral data used in this present case.
11.3 Extent of Data Used in the Study
a) Crop Yield Estimation Survey Data- In this study the yield data for the year 1995-
96 from yield estimation surveys based on crop cutting experiment for wheat crop in
Rohtak district in Haryana has been used. A stratified multistage random is adopted in
these surveys where the tehsil constitute the strata, the villages selected randomly
formed the primary sampling unit, the fields selected from each village formed the
secondary stage unit and the plot within the field formed the ultimate stage of
sampling. A total sample of 75 villages was selected, two fields were selected
randomly and from each field, a plot measuring 5m x 5m was selected for recording
the yield by actual harvesting the crop.
b) Satellite Data- The IRS-1B satellite data was used for recording the yield. The digital
image processing was carried out in the Remote Sensing Laboratory with the help of
Digital image processor software ERMAPPER.
A GPS was also used to identify the exact locations of the plots selected for crop
cutting experiment for wheat crop in terms of their latitudes/longitudes and also the
location of ground control points (GCP’s) which were later used to identify the raw
digital spectral data.
11.4 Method of Estimation
Crop yield estimation surveys based on crop cutting experiments are conducted throughout
the country for obtaining precise estimation of average yield for all major crops.
Here in this case the estimation of average crop yield at district level was done by using crop
yield data obtained from general crop yield estimation surveys based on crop cutting
experiments and the satellite spectral data.
III.190
11.4 Results and Discussion
Table 3: Wheat crop yield estimation using stratification based on NDVI and RVI
for district Rohtak (Haryana), 1995-96
III.191
SPATIAL DATA QUALITY
Tauqueer Ahmad
I.A.S.R.I., New Delhi -110012
1. Introduction
GIS developers and users have paid little attention to the problems caused by error,
inaccuracy and imprecision in spatial data sets. There was awareness that all data suffers
from inaccuracy and imprecision, but effects on GIS problems and solutions were not
considered. It is now generally recognized that error, inaccuracy and imprecision can “make
or break” GIS projects-making the results of a GIS analysis worthless. Spatial analyses done
manually can easily align map boundaries to overlap and be registered. An automated GIS
cannot do this, unless it is programmed to recognize the “undershoots, overshoots, and
slivers” to connect lines. The level of the data quality must be made clear for the GIS to
operate correctly. Assessing the quality of the data, however, may be costly. Data quality
generally refers to the relative accuracy and precision of a particular GIS database. Error
encompasses both the imprecision of data and its inaccuracies.
Although the term "garbage in, garbage out" certainly applies to GIS data, there are other
important data quality issues besides the input data that need to be considered.
2. Components of Data Quality
There are three main components of data quality. (i) Micro level components (ii) Macro level
components (iii) Usage components.
2.1 Micro Level Components
Micro level components are data quality factors that pertain to the individual data elements.
These components are usually evaluated by statistical testing of the data product against an
independent source of higher quality information. They include positional accuracy, attribute
accuracy and logical consistency given as follows:
1. Position Accuracy
2. Attribute Accuracy
3. Logical Consistency
(a) Position Accuracy
Position accuracy is the expected deviance in the geographical location of an object in the
data set (e.g. on a map) from its true ground position. Selecting a specified sample of points
in a prescribed manner and comparing the position coordinates with an independent and more
accurate source of information usually test it. There are two components to position accuracy:
the bias and the precision.
(b) Attribute Accuracy
Attributes may be discrete or continuous variables. A discrete variable can take on only a
finite number of values whereas a continuous variable can take on any number of values.
Categories like land use class, vegetation type, or administrative area are discrete variables.
They are, in effect, ordered categories where the order indicates the hierarchy of the attribute.
(c) Logical Consistency
Logical consistency refers to how well logical relations among data elements are maintained.
It also refers to the fidelity of relationships encoded in the database, they may refer to the
geometric structure of the data model (e.g. topologic consistency) or to the encoded attribute
information e.g. semantic consistency).
2.2 Macro Level Components
Macro level components of data quality pertain to the data set as a whole. They are not
generally amenable to testing but instead are evaluated by judgment (in the case of
completeness) or by reporting information about the data, such as the acquisition date. Three
major macro level components are:
1. Completeness
2. Time
3. Lineage
(a) Completeness
Completeness refers to the exhaustiveness of the information in terms of spatial and attribute
properties encoded in the database. It may include information regarding feature selection
criteria, definition and mapping rules and the deviations from them. The tests on spatial
completeness may be obtained from topological test used for logical consistency whereas the
test for attribute completeness is done by comparison of a master list of geo-codes to the
codes actually appearing in the database.
There are several aspects to completeness as it pertains to data quality. They are grouped here
into three categories: completeness of coverage, classification and verification.
The completeness of coverage is the proportion of data available for the area of interest.
Completeness of classification is an assessment of how well the chosen classification is able
to represent the data. For a classification to be complete it should be exhaustive, that is it
should be possible to encode all data at the selected level of detail.
Completeness of verification refers to the amount and distribution of field measurements or
other independent sources of information that were used to develop the data.
(b) Time
Time is a critical factor in case of any type of data. Some data will be significantly biased
depending on the time period over which they are collected.
Example:
Demographic information is usually very time sensitive. It can change significantly over a
year. Land cover will change quickly in an area of rapid urbanization.
(c) Lineage
The lineage of a data set is its history, the source data and processing steps used to produce it.
The source data may include transaction records, field notes etc. Ideally, some indication of
lineage should be included with the data set since the internal documents are rarely available
and usually require considerable expertise to evaluate. Unfortunately, lineage information
most often exists as the personal experience of a few staff members and is not readily
available to most users.
III.194
2.3 Usage Components
The usage components of data quality are specific to the resources of the organization. The
effect of data cost, for example, depends on the financial resources of the organization. A
given data set may be too expensive for one organization and be considered inexpensive by
another.
Accessibility refers to the ease of obtaining and using the data. The accessibility of a data set
may be restricted because the data are privately held. Access to government-held information
may be restricted for reasons of national security or to protect citizen rights. Census data are
usually restricted in this way. Even when the right to use restricted data can be obtained, the
time and effort needed to actually receive the information may reduce its overall suitability.
The direct cost of a data set purchased from another organization is usually well known: it is
the price paid for the data. However, when the data are generated within the organization, the
true cost may be unknown. Assessing the true cost of these data is usually difficult because
the services and equipment used in their production support other activities as well.
The indirect costs include all the time and materials used to make use of the data. When data
are purchased from another organization, the indirect costs may actually be more significant
than the direct ones.
It may take longer for staff to handle data with which they are unfamiliar, or the data may not
be compatible with the other data sets to be used.
3. Causes of Error
In this section, it is examined when and how errors creep into GIS data. The three major
causes of GIS data error are problems found in (i) source data, (ii) data entry and (iii) data
analysis.
3.1 Errors in Source Data
It has become common now to collect GIS data directly in the field. Data collection can be
done using field survey instruments that download data directly into GIS or via GPS receivers
that directly interface with GIS software on portable PCs. These techniques can eliminate the
need for GIS source data.
But during the last many years, GIS data most often have been digitized from several sources,
including hard copy maps, rectified aerial photography and satellite imagery. Hard-copy
maps (e.g. paper, vellum and plastic film) may contain unintended production errors as well
as unavoidable or even intended errors in presentation. The following are "errors" commonly
found in maps.
Map Generalization
Cartographers often deliberately misrepresent map features due to limitations encountered
when working at given map scales. Complex area features such as industrial buildings may
have to be represented as simple shapes. Linear features such as roads may have to be
represented by parallel lines that appear wider on a map. Curvilinear features such as streams
may have to be represented without their smaller twists and bends.
Indistinct Boundaries
Indistinct boundaries typically include the borders of vegetated areas, soil types, wetlands and
land use areas. In the real world, such features are characterized by gradual change, but
cartographers represent these boundaries with a distinct line. Some compromise is inevitable.
III.195
Map Scale
Cartographers and photogrammetrists work to accepted levels of accuracy for a given map
scale as per National Map Accuracy Standards. Locations of map features may disagree with
actual ground locations, although the error likely will fall within specified tolerances. Of
course, the problem is compounded by limitations in linear map measurements-typically
about 1/100th of an inch on a map scale.
Map Symbology
It is impossible to perfectly depict the real world using lines, colors, symbols and patterns.
Cartographers work with certain accepted conventions. As a result, facts and features
represented on maps often must be interpreted or interpolated, which can produce errors. For
example, terrain elevations typically are depicted using topographic contour lines and spot
elevations. Elevations of the ground between the lines and spots must be interpolated. Also,
areas symbolized as "forest" may not depict all open areas among the trees.
3.2 Errors during Data Entry
GIS data typically are created from hard copy source data. The process often is called
"digitization", because the source data are converted to a computerized (digital) format.
Human digitization can compound errors in source data as well as introduce new errors. The
following are the primary methods of digitizing hard copy source data:
Manual Digitizing
Although manual digitizing is used less often today, it was the predominant digitizing method
in the 1980s. Maps are affixed to digitizing tables, registered to a GIS coordinate system and
"traced" into a GIS. A digitizing table has embedded in its surface a fine grid of wires that
sense the position of a cross hair on a hand held cursor. When a cursor button is pressed, the
system records a point at that location in the GIS database. The operator also identifies the
type of feature being digitized as well as its attributes.
Photogrammetric mapping also is a manual digitizing process. Through an exacting and
rigorous technical process of aerotriangulation, overlapping pairs of aerial photographs are
registered to one another and viewed as a 3-D image in a stereoplotter or via special 3-D
viewers. In a process called "stereocompilation," a photogrammetrist traces map features that
are encoded directly into a database.
Scanning and Keyed Data Entry
In scanning, source data are mechanically read by a device that resembles a large format copy
machine. Sensors encode the image as a large array of dots, much like a fax machine scans a
letter. High resolution scanners can capture data at 2,000 dots per inch (dpi), but maps and
drawings typically are scanned at 100 dpi to 400 dpi. The resulting raster image then is
processed and displayed on a computer screen. Further onscreen manual digitizing (i.e.
"heads-up digitizing") usually is needed to complete the data entry process. If the source data
contain coordinate values for points or the bearings and distances of lines (e.g. parcel lines),
then map features can be keyed into a GIS with great precision.
General Data Entry
Accurate digitizing is not easy. It requires certain basic physical and visual skills as well as
training, patience and concentration. There also are many opportunities for error, because the
process is subject to visual and mental mistakes, fatigue, distraction and involuntary muscle
movements. In addition, the "set up" of a map on a digitizing table or a scanned raster image
III.196
can produce errors. Cell size of a scanned raster image also can affect the accuracy of heads-
up digitizing.
A digitizer must accurately discern the center of a line or point as well as accurately trace it
with a cursor. This task is especially prone to error if the map scale is small and the lines or
symbols are relatively thick or large. The method of digitizing curvilinear lines also affects
accuracy. "Point-mode" digitizing, for example, places sample points at selected locations
along a line to best represent it in a GIS. The process is subject to judgment of the digitizer
who selects the number and placement of data points. "Stream-mode" digitizing collects data
points at a pre-set frequency, usually specified as the distance or time between data points.
Every time an operator strays from an intended line, a point digitized at that moment would
be inaccurate. This method also collects more data points than may be needed to faithfully
represent a map feature. Therefore, post-processing techniques often are used to "weed out"
unneeded data points.
Heads-up digitizing often is preferred over table digitizing, because it typically yields better
results more efficiently. Keyed data entry of land parcel data is the most precise method.
Moreover, most errors are fairly obvious, because the source data usually are carefully
computed and thoroughly checked. Most keyed data entry errors show as obvious
mismatches in the parcel "fabric."
GIS software usually includes functions that detect several types of database errors. These
error-checking routines can find mistakes in data topology, including gaps, overshoots,
dangling lines and unclosed polygons. An operator sets tolerances that the routine uses to
search for errors, and system effectiveness depends on setting correct tolerances. For
example, tolerances too small may pass over unintentional gaps, and tolerances too large may
improperly remove short dangling lines or small polygons that were intentionally digitized.
3.3 Errors during Data Analysis
Even if "accurate," the manipulation and analysis of GIS data can create errors introduced
within the data or produced when the data are displayed on screen or plotted in hard copy
format.
4. Sources of Possible Errors
4.1 Obvious Sources of Error
1. Age of data. With the exception of geological data the reliability decreases with
age.
2. Areal coverage: partial or complete. Many countries still have fragmentary
coverage of maps at scales of 1:25000 to 1:50000.
Moreover, during the last 30 to 40 years, concepts and definitions of map units,
the way they should be mapped have changed.
3. Map scale.
4. Density of observation. How dense should observations be "to support a map".
5. Relevance. Not all data used are directly relevant for the purpose for which they
are used. Prime example: remotely sensed data.
6. Accessibility. Not all data are equally accessible (e.g. military secrecy).
Sometimes data is not available because of inter-department secrecy.
7. Cost.
III.197
4.2 Errors Resulting from Natural Variations or from Original Measurement
1. Positional accuracy. Topographical data often available with a high degree of
positional accuracy. But position of vegetation boundaries etc. often influenced by
the subjective judgment of surveyor or by interpretation of remotely sensed data.
2. Accuracy of content: qualitative and quantitative. The problem of whether the
attributes attached to points, lines or polygons are correct and free from bias.
Sometimes systematic errors occur because of instrument. If pixel is too large then
it is not clear that it should be classified as forest, road or camp.
3. Sources of variations in data: data entry, observer bias, natural variation. Data
entry error. Field data very much influenced by surveyor (elevations or census
takers).
4. Errors arising through processing. Numerical errors in the computer (e.g. joining,
matching of a field in GIS software), rounding off errors, truncation etc.
4.3 Numerical Errors in Computers
1. Faults arising through topological analyses. Problems associated with map
overlay. Digitizing considered infallible. Boundary data is assumed to be sharply
defined. All algorithms are assumed to operate in a fully deterministic way.
2. Classification and generalization problems: class intervals, interpolation.
The concepts of complete, compatible, consistent and applicable GIS data previously defined
particularly apply to data analysis. Users must consider whether a selected GIS dataset is
complete, consistent and applicable for an intended use, and whether it is compatible with
other datasets used in the analysis.
The phrasing of spatial and attribute queries also may lead to errors. In addition, the use of
Boolean operators can be complicated, and results can be decidedly different, depending on
how a data query is structured or a series of queries are executed. For example, the query,
"Find all structures within the 100 year flood zone," yields a different result than, "Find all
structures touching the 100 year flood zone." The former question will find only those
structures entirely within the flood zone, whereas the latter also will include structures that
are partially within the zone.
Dataset overlay is a powerful and commonly used GIS tool, but it can yield inaccurate
results. To determine areas suitable for a specific type of land development project, one may
overlay several data layers, including natural resources, wetlands, flood zones, land uses, land
ownership and zoning. The result usually will narrow the possible choices down to a few
parcels that would be investigated more carefully to make a final choice. The final result of
the analysis will reflect any errors in the original GIS data. Its accuracy only will be as good
as the least accurate GIS dataset used in the analysis.
It is also common to overlay and merge GIS data to form new layers. In certain
circumstances, this process introduces a new type of error: the polygon "sliver." Slivers often
appear when two GIS datasets with common boundary lines are merged. If the common
elements have been digitized separately, the usual result will be sliver polygons. Most GIS
software products offer routines that can find and fix such errors, but users must be careful in
setting search and correction tolerances.
III.198
5. Controlling Errors
GIS data errors are almost inevitable, but their negative effects can be kept to a minimum.
Knowing the types and causes of GIS data errors is half the battle; the other half is employing
proven techniques for quality control at key stages in the GIS "work flow."
Many errors can be avoided through proper selection and "scrubbing" of source data before
they are digitized. Data scrubbing includes organizing, reviewing and preparing the source
materials to be digitized. The data should be clean, legible and free of ambiguity. "Owners"
of source data should be consulted as needed to clear up questions that arise.
Data entry procedures should be thoroughly planned, organized and managed to produce
consistent, repeatable results. Nonetheless, a thorough, disciplined quality review and
revision process also is needed to catch and eliminate data entry errors. All production and
quality control procedures should be documented, and all personnel should be trained in these
procedures. Moreover, the work itself should be documented, including a record of what was
done, who did it, when was it done, who checked it, what errors were found and how they
were corrected.
To avoid misusing GIS data and the misapplication of analytical software, GIS analysts
including casual users need proper training. Moreover, GIS data should not be provided
without metadata indicating the source, accuracy and specifics of how the data were entered.
References
1. Burrough, P.A. (1986). Principles of Geographic Information Systems for Land
Resource Assessment. Clarendon Press, Oxford.
2. Cressie, N. (1991). Statistics for Spatial data. John Wiley and Sons Ltd., New York.
3. Bernhardsen, T. (2002). Geographic Information Systems: An Introduction. John
Wiley & Sons, Inc.
4. Burrough, P.A. and McDonnell R.A. (1986). Principles of Geographic Information
Systems. Oxford: Oxford University Press.
5. Buckley, D.J. The GIS Primer: An Introduction to Geographic Information System.
http://www.innovativegis.com/basis/primer/primer.html.
III.199
APPLICATION OF GIS AND REMOTE SENSING
IN LAND USE STATISTICS
Anil Rai
I.A.S.R.I., New Delhi -110012
1. Introduction
Systematic and comprehensive compilation of land use statistics is necessary for planned
development of agriculture, forests, grasslands, rural settlements, urban spreads, industries
and other land based programmes and activities. The need for optimizing land use in an
integrated manner has become particularly relevant in recent years as a consequence of
compelling and conflicting demands of growing population, increasing land degradation and
thus sharply declining man land ratio. The land use planning is a means of optimum
utilization of natural resources, specially, the land which is one of the most important
resource, based on its related factors like socio-economic condition, soil type, climate, water
availability etc. for production as a whole. Naturally, the effective land use planning of
particular region needs information related to various input factors of production.
The land use statistics not only helps in this planning process but also generates information
about other related factors. In India the land use statistics are obtained as per nine fold
classification, which is vogue since 1950-51.These statistics are complied and published by
Directorate of Economics and Statistics(DES) mostly at district level for most of the states in
the country, where cadastral maps exist by complete enumeration through primary reporting
agency called Patwari. In permanently settled states of Kerala, Orissa and West Bengal these
statistics are based on sample surveys covering 20% of the villages every year.
It is general feeling that due to cost and organizational constraints of agencies engaged in
collection of data on this aspect, the quality of data is deteriorating and also there is
considerable delay in bringing out the results, which hampers the efficient planning
process.Further, in view of the micro level planning, there is greater need to provide these
statistics at lower levels like tehsils, blocks or at gram panchayat level (as in case of National
Crop Insurance Scheme)
The recent advances in the field of geographical techniques, like Geographic Information
System (GIS) and Remote Sensing (RS) increased the potential to change substantially the
statistical approach to study the geographical realities. The remote sensing techniques to
obtain the land utilization statistics gained popularity because of its extensive coverage of
geographical area. Although, this technology has vast potential especially in this area, but
statistics so obtained are also not free from various kinds of errors. The GIS, is capable of
handling geographical data through its geographical coordinates, obtained by different
sources like Census, Survey and Remote Sensing. It is one of the important tool having
capabilities of integration of different kinds of data and assist in the spatial analysis of data
from geographical characters. These recent technologies i.e. GIS & Remote Sensing
combined together named as geomatics, can be used as a tool to resolve some of the above
issues affecting the land use statistics.
Recently, National Statistical Commission (NSC) reviewed the National Agricultural
Statistical System in a great depth and made a important suggestions related to improvement
of in the Agricultural Statistical System and quality of data. In relation to crop area statistics,
NSC suggested that a 20% sample of villages is large enough to estimate crop area quite
efficiently at the state and district level and hence crop area estimation should be based on
20% of sampled villages.
This study was formulated to explore the various possibilities of obtaining reliable land
utilization statistics with minimum time lag by considering these important sources of data
collection and processing. The study aimed to (i) examine the quality of data on land use
statistics recorded by the village agency (Patwari) by independently observing a sample of
villages by specially trained and qualified investigators. (ii) Study also aimed to improve the
efficiency of the estimator based on traditional sample surveys by using spatial properties of
land use and (iii) spatial models were fitted for correcting the estimates of land use statistics
obtained through usual approach of satellite digital data. Further, qualitative aspects of data
from revenue records have been studied using survey data as true value. The sampling units
in the sample i.e. villages have been selected with the help of GIS by incorporating spatial
correlation in the sampling design it self. The spatial models were developed using Remote
Sensing data as independent variable and exploiting the properties of spatial relationship of
different classes of land for prediction of area under different land use characters. With the
help of these spatial models, the area statistics can be easily obtained at lower geographical
level under different land use classes in a village at highly reduced cost and with the help of
revenue records or field surveys. The study was undertaken in district of Lalitpur in UP as
this district has been observed to have considerable area under most of the land use
classification categories. The study was initiated with following objectives:
(1) To obtain land utilization statistics with the help of survey and Remote Sensing
technique.
(2) To study the qualitative aspects of land utilization statistics obtained through different
sources i.e. Census, Survey and Remote Sensing.
(3) To develop model for integration of statistics obtained through different sources.
IV.201
studies have been carried out to develop methodologies for application of satellite data in
various fields of agricultural and rural development by Department of Space. Some of these
studies are by Dadhwal et al. (1985, 1991), etc. The details of GIS and remote sensing
techniques are provided by Burrough (1986), Curan (1985), Goodchild, (1987). Several
methodological studies related to estimation of crop area and production have been carried
out at Indian Agricultural Statistics Research Institute (IASRI), New Delhi. Singh et al.
(1992) used satellite data for stratification of crop area for the general crop estimation surveys
and obtained more precise estimator of crop yield. Singh et al. (1999) also developed small
area estimator of crop yield. Singh et al. (2002) used satellite data and the farmers eye
estimate for developing a reliable crop yield model.
3. Data Collection
In this study, data from three different sources namely revenue records based on complete
enumeration of villages; sample survey and remote sensing have been used as per the
objectives of the study. In order to plan the field study of district, the census data of 1991 has
been obtained from the Office of Registrar General of India, New Delhi. This data was in
digital format and it has been utilized for planning of the field work and selection of sample.
A sample of size 20 villages has been selected from the district following proportional
representation of the villages of each tehsils.The copy of revenue records collected by the
respective patwaris of the selected villages for all three seasons in the district has been taken
from the tehsil head quarters. This record pertains to survey number wise land use statistics
under different classes for respective villages. In order to check the quality of the data
collected by the patwaris in the selected villages, field staff of the institute was deputed to
completely enumerate all survey numbers of the selected villages using categories of the area
falling under different land use classes as per the standard nine fold classification. The
schedules designed for this data collection were having provision of collection of all
information, which were collected by the patwaris apart from other parameters. The copy of
cadastral maps of the selected villages have been obtained from tehsils head quarter and
provided to field staff to assist in data collection. The data collected by two independent
sources on same set of 20 villages were used for qualitative checking of revenue records. The
total numbers of survey numbers covered in each season under this study were 36320
Further, satellite remote sensing data from Indian Remote Sensing Satellite IRS-1C, LISS-III
sensor as well as WiFs for all three seasons has been procured from NRSA, Hyderabad.
Further, training sites belonging to different land use classes were identified in all three
seasons with the help ground survey carried out in the entire district. In identification of these
training sites, Survey of India toposheets of 1:50,000 and 1: 2,50,000 have been extensively
utilized.
4. Spatial Stratified Sampling Technique
In present study a stratified spatial sampling technique has been suggested for selection of
areal sampling units (villages) in space from the district (spatial region), which is a function
of combined spatial correlation coefficient for different land use variables and geographical
area (i.e. size) of the villages. Tehsils (lower administrative units in a district) in a districts
forms the strata and villages in a tehsil form the sampling units. The neighbors are defined on
the basis of optimum distances, from centriod of each village. In this study census data for
1991 census for district Lalitpur has been used for sample selection. There are 754 villages in
the district and three land use variables which are available in the census data with respect to
land use namely (i) Un-irrigated area (U) (ii) Culturalble waste (W) and (iii) Area not
available for cultivation (C) have been used for sample selection.
IV.202
Let X / X1 , X 2 ,......., X N a vector of a land use variable deviated from its mean. The
corresponding lagged variable as defined by Cliff and Ord (1981) is given by
/
X* W * X
where
0 w12 w1N
w 0 w2 N
21
W
*
wN 1 wN 2 0
X / X
1
2
X XX X
*/ * /
1
2
… (1)
In the present study spatial correlations for different land use variables i.e. unirrigated area
( U ), culturable waste ( w ), and area not available for cultivation ( c ) were calculated.
Let their variances be denoted by u2 , w2 and c2 respectively. Since, number of land use
variables needs to be estimated through same sample, so it is desirable to estimate a
combined spatial correlation coefficient for selection of the sample. A reasonable value of
combined correlation coefficient can be obtained as
u2 u w2 w c2 c
… (2)
u2 w2 c2
This combined spatial correlation was used for defining neighborhoods, so that this
sampling design is reasonably efficient for all land use classes in the region. In this study
distance based neighbors were defined by associating information at the centroid of each
village and obtaining optimum distance from the village under consideration. The centroid
is the point with coordinates that are the mean values of coordinates of all points in the
configuration (polygon). The distances between any two areal units (villages) were taken to
be Euclidean distance between the centriods of areal units (villages).
5. Optimum Distance for Defining Neighbors
Once distance between any two villages was computed, the next step was to decide areal
extent of neighborhood. A circle of radius r is considered from the centriod of the village
for which the neighbors were to be defined. All those villages whose centriods fall within
this radius were identified as first lag neighbors. Villages whose centroids were falling were
identified as second lag neighbors. Similarly, third, fourth lag neighbors were identified.
IV.203
The optimum value of r was obtained based on data from census 1991 on 754 villages of the
district. To obtain the optimum value of r, the spatial correlation for different land use
variables along with their variances were calculated for different distances (in degree
decimals) h. It was observed that an increase in spatial correlation occurs with increase in
values of h up to certain point and after that spatial correlation started decreasing. The
variance was also minimized at this peak value of for a particular value of h, which was
considered as the optimum distance (or radius) for defining neighbors. The optimum value
of distance i.e. ro was defined as the weighted average of optimum distances obtained for
the three land use variables as given below
u2 hu w2 hw c2 hc
r0 … (3)
u2 w2 c2
After obtaining value of combined spatial correlation coefficient and defining criterion
for neighbors using optimum distances, a stratified sample of size n villages from a
population of size N villages were selected by using proportional allocation following a
spatial sampling procedure, which incorporates spatial correlation coefficient and size
measure of areal units with respect to neighbors of already selected units in the sample. The
spatial sampling procedure is given in the following section.
Sample Selection: Let the population is (a district) divided in L strata (Tehsil). Let Nl be
the number of sampling units ( Village) in the l-th strata. Let nl denote the sample size of
villages to be selected from the l th strata. Following steps were followed for selecting
sampling units from a stratum:
Step 1: Selection of first unit from a stratum
The first unit in each stratum was selected with probability proportional to size measure i.e.
geographical area of the village X g . Let X gil denote the geographical area of the ith village
in the lth stratum and X gl denotes total geographical area of the lth stratum. Then the
probability of selecting ith unit (village) of lth stratum at first draw in the sample is P(1)il =
X gil
.
Xgl
IV.204
4. Select the unit i if j U i2 (l ) X gil .
5. Reject unit i and repeat the above process if j U i2 (l ) X gil .
The probability of selecting ith unit of the pupulation in the 2nd draw is given by
U i2 (l ) X gil
P( 2)i l
U i2 (l ) X gil
il
isl*
Here, l denotes the set of all units belonging to stratum l and s l* denotes the set of earlier
selected units in the sample. It can be seen that sum of these probabilities over N l 1 units
of the population becomes one.
Step 3: Selection of subsequent Units
U i3 (l ) = 1
d( l )13
1 d( l ) 23
U i4 (l ) = 1
d ( l )14
1 d ( l ) 24
1 d ( l ) 34
…………….
…………….
…………….
U in
l
(l ) = 1 d ( l )1nl
1 d ( l ) 2 nl
... 1 d ( l ) nl 1nl
It can be easily seen that the probability of selecting ith unit in n l -th draw is given by
U in (l ) X gil
Pinl (l ) Nl
l
.
U inl (l ) X gil
il
isn*l 1
IV.205
denotes the area under a th land use category from ith village of lth stratum. Now sample
mean of lth stratum for ath category i.e. y a (l ) was obtained using following estimator
nl d ai(l )
y a (l )
i 1 nl
where
y a1(l )
d a1(l )
N l P(1)il
1 y a 2(l )
d a 2(l ) y a1(l )
Nl P( 2)il
……………
……………
……………
1 y anl (l )
d an(l ) y a1(l ) y a 2(l ) . . .
Nl P( nl )il
Then the estimator for area under ath land use category in a district was obtained by
L
Ya N Wl y a (l ) … (4)
l 1
IV.206
7. Spatial Regression Model
Let, Ya denotes the vector of observations pertaining to the values of ath class of land use
through a suitable model is developed using this data of villages selected in the sample.
Similarly X a is corresponding vector of observation obtained through patwari records.
Further as NSC has recommended to do away with the complete girdavali and use only data
for 20% villages for accarage estimation, this methodology of spatial modeling can be used
with greates advantage to improve the land use statistics using the patwari land records of this
20% sample of villages and the satellite data of the whole district of all the villages.
Let us consider the following simple regression model
Y Xβ e
where e ~ N 0, 2 I
1
Y1 Y1,a
Y Y 2
Y ,
2
Ya
2,a
, β , (a = 1,2,3,…,9)
Y20,a
Y9 9
X1 02 09
0 X2 09
1
X 01 02 09 and I is identity matrix.
01 02 09
01 02 X 9
In this model errors e are assumed to follow normal distribution with mean vector 0 and
variance-covariance matrix σ 2 I . So ordinary least square estimator of β i.e ̂ ols is identical
to its maximum likelihood estimator. It is well known that in case of spatial data the structure
of variance – covariance matrix of error e i.e. σ 2 I will not have simplified structure due to
spatial dependence of observations. Hence, structure of variance –covariance matrix here
may be represented by . The elements of may be obtained by following two different
approaches.
8. Spatial Weighted Regression Model
This approach is based on obtaining optimum weights depending on the location of the
spatial aerial units. Hence, in this approach the location specific regression coefficient can be
obtained. This technique provides non-parametric estimates of regression coefficients. It also
leads to localized ordinary least square estimation and may be called spatial weighted
regression. This technique was suggested by Brunsdon, Furtheringham & Chariton(1998).
IV.207
In this case for a given location “i” the circle of inclusion was drawn using the optimum
radius r0 obtained for defining the neighbors. The weights i k for k –th villages were
obtained using the procedure as given below:
exp( d ik / s r0 ), if d ik s r0
ik … (6)
0, otherwise
Where, s denotes the order of neighborhood and takes values 1,2, 3,….and dik is the distance
between i-th and k-th location. This is also called Kernel function or Kernel and denoted by
the letter K such that i k = K(dik). The above Kernel has all the desirable feature of a Kernel
function such as
(i) K (0) =1
1 a12 a1n
1 a 2 n
a 21
2
Va= a a 31 a 32 1 a 3n a 1,2,3,..., 9
1
an1 an2 1
Thus using this structure of , location specific equations were obtained separately. Ya ,
area under different land use classes can be obtained using the estimated regression
coefficients.
̂ X X 1 X Y
The estimate of variance for the above estimator of regression coefficient is
Vˆ X 1 X
1
IV.208
realistic fashion, the possible spatial dependence of the errors. The simplest way of denoting
the spatial dependence of errors is the use of authorized variogram function. In this approach,
elements of Va i.e. ij are obtained by using following equations:
r h 1 e h / r0
where r0 is optimum distance parameter and h is the lag distance.
The spherical variogram function is defined as
3h 1 h
, h r0
r (h) 2r0 2 r0
1 , h r0
The variance–covariance structure can be estimated based on these two veriogram
functions spatially. The Gausian variogram function was also fitted but the fit was poor so it
has not be considered further. Using the data of selected villages in these models, the land use
statistics of the non-selected villages can be obtained.Again the estimated variance-
covariance for the regression coefficient is given by
Vˆ X 1 X
1
IV.209
The best-fitted model for each class may be different in different regions depending on the
distribution pattern of the land use categories. The comparison of the estimates of area under
different classes of land use obtained from field surveys, spatial models and remote sensing
has been also presented. These estimates were obtained for only those classes which were
identifiable in single year and single date IRS-IC satellite data from LISS-III sensor. It may
be noted that in case of remote sensing, one year data, only five classes can be identified out
of nine classes of traditional nine fold classification of land use. The land under forest (Class-
1), Barren and unculturable land (Class-2) and land put to non-agricultural use (Class-3) are
identified clearly using digital data. The area under different crops (Class-5) is mixed with
the area under permanent pastures and other grazing lands (Class-4) and area under net area
sown (Class-9). Another new class is formed as fellow land, which is a combination of area
under cultivable waste (Class-6), area under current fellow (Class-7) and area under fellow
land other than current fellow (Class-8). The variance and % standard errors for different
land use classes were also calculated only for field survey data by incorporating sampling
design for selection of the sample. The salient achievements of this analysis are as follows:
1. The data quality of revenue records were found to be quite satisfactory at least for
nine fold classification of land use classes as compared to field survey data in the
study district. Hence, to obtain land use statistics falling under each of the classes, the
data of revenue records could be used even for sampled villages instead of completely
enumerating these villages.
2. The prediction of area under different land use categories covered under nine-fold
classification based on satellite data using spatial model seems to be quite
satisfactory. In case of land use class1and 2 accuracy of prediction seems to be poor.
The best-fitted model for each class may be different in different regions depending
on the distribution pattern of the land use categories. It may be noted that these results
holds true only for the study district i.e. Lalitpur. This analysis for development of the
spatial models has to be performed for each district/region separately as the spatial
pattern of land use classes for other district/region may be different.
3. The estimates of area under different land use classes obtained through remote
sensing have been corrected through spatial models. The overall correction factor is
around 4% of the total geographical area with respect to actual geographical area of
the district.
4. The spatial models developed in this project may be used for estimation of area under
different land use statistics even at village level. Hence, these models are quite useful
for obtaining these land use statistics at small area levels, such as village
panchayats/blocks/tehsils etc..
5. It is observed that the techniques developed under this project may be useful for
implementation of National Statistical Commission recommendations in which it has
been observed that these statistics should be obtained only on the basis of 20%
sampled villages. It may be noted that by remote sensing technology and spatial
models, the sample size of the villages may be further reduced, which may leads to
considerable saving of national resources.
The total of estimated area for different land use classes using spatial model is around
14% away from actual geographical area where as the total area of all the classes
obtained through remote sensing is almost 18% higher than actual geographical area
of the district. It clearly indicates that the overall correction of area estimates obtained
through remote sensing is around 4% by using spatial models. It may be noted that the
total estimates of geographical area obtained through spatial survey is closer to actual
IV.210
geographical area of the district, it does not imply that the estimates of area under
different land use classes estimated through survey are also close to actual area under
the same class in the district. Further, it is not possible to estimate area at small area
level like village panchayats/blocks/tehsils using spatial sampling technique whereas
spatial models will provide area estimates at any of these small area levels.
References
1. Arbia, G. (1993). The use of GIS in spatial statistical surveys. Int. Statist. Rev., 61(2),
339- 359.
2. Cliff, A.D. and Ord, J.K. (1981). Spatial Processes, PION, London.
3. Des Raj (1956). Some estimators in sampling with varying probabilities without
replacement. J. Amer. Statist. Assoc., 51, 269-284.
4. Misra, P. (2001). Application of Spatial Statistics in Agricultural Surveys. Ph.D.
Thesis, IASRI, New Delhi.
IV.211
CROP ESTIMATION AND FORECASTING
USING REMOTE SENSING
Randhir Singh
1. Introduction
Agriculture is the backbone of Indian economy, contributing about 40 percent towards the
Gross National Product (GNP) and providing livelihood to about 70 percent of the
population. So for a primarily agriculture based country like India, reliable, accurate and
timely information on types of crops grown and their acreages, crop yield and crop growth
conditions are of vital importance. India is one of the few countries which has a well-
established system of collection of agricultural statistics and detailed statistics of land
utilization are continuously available since 1884. The agricultural crop production of
principal agricultural crops in the country is usually estimated as a product of area under the
crop and the average yield per unit area of the crop. The estimates of the crop acreage at a
district level are obtained through complete enumeration whereas the average yield is
obtained through general crop estimation surveys (GCES) based on crop cutting experiments
conducted on a number of randomly selected fields in a sample of villages in the district.
The traditional approach of crop estimation in India involves complete enumeration (except a
few states where sample surveys are employed) for estimating crop acreages and sample
surveys based on crop cutting experiments for estimating crop yield. The crop production
estimates are obtained by taking product of crop acreage and the corresponding crop yield.
The yield surveys are fairly extensive with plot yield data collected under a complex
sampling design that is based on a stratified multistage random sampling design. In each
district, tehsils/blocks form the strata, the villages in a block form the primary sampling units,
the fields under the crop form second stage unit and at the third stage of sampling a plot of
specified dimensions within a field is selected for harvesting to determine the crop yield. The
sample units are randomly selected at each stage of sampling. For a detailed description of
the methodology of data collection in crop surveys one may refer to Sukhatme and Panse
(1951).
These crop surveys are planned with a specific population (district) in view. However in
view of growing needs of micro level planning and especially the demand of crop insurance,
the need for reliable small area crop statistics for sub populations (like Tehsil or Block) is
increasing. But planning of surveys for Tehsil level estimates will require much larger
number of crop cutting experiments and the costs involved will be quite prohibitive. The
advances in remote sensing technology applications and also the advances in computing
facilities have provided some convenient techniques of small area estimation which borrow
strength from related or similar larger area through explicit and implicit models that connect
the small area via supplementary data. Hence estimates at Tehsil level can be attempted
through the crop surveys earlier planned for district level estimators.
The crop forecasts/advanced estimates of crops are presently developed by the Ministry of
Agriculture. The advance estimates of kharif crops are first prepared in July/August
tentatively when behavior of South West monsoon is clear and reports of coverage of area
under crops from the states are available. The advance estimates are reviewed during
December/January when estimates of area under kharif crop become available under the
Timely Reporting Scheme (TRS) and results of the crop cutting experiments portion from the
NSSO (normally 10%) become available. The advance estimates of rabi season are also
212
prepared at this stage. The advance estimates are again reviewed in the month of April based
on information obtained from the states giving the final forecast for kharif.
With the advent of Remote Sensing Technology during 1970s, its great potential in the field
of agriculture have opened new vistas of improving the agricultural statistics system all over
the world. Space borne remotely sensed spectral satellite data has been widely used in the
field of agriculture for estimation of area under different major crops like wheat, paddy,
groundnut and sugarcane. Studies have also been made to examine the relationship of crop
growth parameters like leaf area index (LAI) representing crop vigour and the spectral data in
the form of several vegetation indices developed from the spectral data of various bands.
Remote sensing satellite data can also be used for improving the crop yield estimation
through crop cutting experiments and also for developing models for crop yield using
historical data, meteorological data along with the remotely sensed satellite data.
During 1990-93 a study was conducted at the Indian Agricultural Statistics Research Institute
(IASRI), New Delhi to examine the usefulness of satellite spectral data for stratification of
crop area based on vegetation indices for improving crop yield estimation based on yield data
from crop cutting experiments under crop yield estimation surveys. The study pertained to
wheat crop yield for district Sultanpur UP for 1985-86 and the satellite data was used from
the USA satellite Land Sat-4. This study showed that the efficiency of crop yield estimation
can be increased considerably by using the satellite data along with the survey data. The
results of this study are given in Singh et al. (1992). Another similar study was undertaken
during 1996-98 for improved estimation of wheat crop yield in district Rohtak for 1995-1996
using the IRS 1B - LISS II satellite data. The results from this study are presented in Singh et
al. (1999) also showed that satellite data in the form of vegetation indices greatly improves
the efficiency of crop yield estimator. The methodology followed in these studies and the
results obtained are presented in the subsequent sections.
2. Crop Yield Estimation using Remote Sensing and Forecasting
Remote sensing technology has been proved as a useful application for natural resources
evaluation and management. In the past, spectral data acquired via satellite have been
extensively utilized for crop yield modeling in various parts of the world. The important
assumption in use of remote sensing data for crop modeling is that the spectral data is
strongly related with canopy parameters which are related to final yield at a critical stage of
the crop growth. Several studies have established good correlation between vegetation indices
and grain yield using single date data near heading/flowering like Colwell et al. (1997),
Serafini (1985), Barnett and Thompson (1982, 83). In India a major project on “Crop
Acreage and Production Estimation (CAPE) has been carried out under the Remote Sensing
Applications Mission by Department of Agriculture and Cooperation and Department of
Space to develop methodology for state level acreage and production estimation of important
crops. Use of satellite spectral data in crop yield estimation surveys was recently examined
by Singh et al (1992). In the study satellite data at the time of maximum flowering/heading
of the crop in the form of vegetation indices was used to stratify the crop area into vegetation
vigour classes like high vegetation, average vegetation, poor vegetation etc. This was
followed by developing certain post stratified estimator of crop yield based on yield data as
obtained from crop cutting experiments. This study showed a substantial improvement in the
efficiency of crop yield estimator as compared to the existing estimator. This suggested that
with the use of satellite spectral data along with the survey data based on crop cutting
experiment either the number of crop cutting experiments can be reduced resulting in
considerable reduction in the cost for obtaining district level estimate with the same precision
IV.213
or small area estimates at a lower level compared to the district (like tehsils or blocks) can be
obtained with the existing crop cutting experiments in a district.
2.1 Estimation Procedure based on use of Crop Yield Survey Data along with Satellite
Spectral Data.
The factors like different soil types, agricultural inputs, adoption of improved technology, etc.
affect the crop yield and hence cause a lot of variability in the yield even within a stratum.
Since the spectral reflectance is a manifestation of all important factors affecting the crop,
hence stratification of crop area on the basis of crop vigour as reflected by the spectral data is
expected to result in a greater efficiency of the crop yield estimation. Singh et al. (1992),
Singh and Goyal (1993) showed that the stratification based on NDVI at heading/flowering
stage improved the efficiency of crop yield estimation considerably.
In the present study it is proposed to develop post stratified estimator of crop yield, for
district Rohtak in Haryana and also to develop small area crop yield estimates at tehsil level
using the crop yield survey data and the satellite spectral data. Crop yield data for 1995-96
obtained from general crop yield estimation surveys and the satellite spectral data of IRS-1B
LISS-II dated Feb. 17, 1996 have been used in the study.
Initially District Boundary Mask was generated using the topographic maps of scale
1:250,000. False Colour Composite (FCC) using spectral data from band-2 (Green) band-3
(Red) and band-4 (Near infra red) was generated. For identifying the villages selected for
crop surveys on the FCC, topographic maps of scale 1:50,000 which contain information on
identifiable features like roads, canals, water bodies etc. were used. Further, a Global
Positioning System (GPS) was used to identify the crop plots. The GPS was taken to the
plots and location of the plots in terms of longitudes and latitudes were recorded. These
locations were then identified on the FCC’s. The coordinates of each plot in terms of
scanline and column number were recorded to identify these plots on the Normalized
Difference Vegetation Index (NDVI) and the Ratio Vegetation Index (RVI) imageries.
NDVI is defined as the ratio of (IR-R) to (IR+R) and RVI is defined as ratio of IR to R where
IR and R denote the radiance in infra red (band 4) and red (band 3) respectively. The concept
of density slicing technique was used to divide the NDVI and RVI imageries into different
identifiable classes which provided three classes named as (i) Non vegetation class (ii)
Average vegetation class and (iii) the High vegetation class. To obtain the strata weights for
the two vegetation classes, the number of pixels falling into these classes were obtained and
since the pixel size is fixed, hence the area of the classes were obtained which were used as
strata weights. Assigning different colours to different class range values, the stratified
imageries were developed.
2.2. Estimation of Crop Yield at District Level
Initially FCC using spectral data from band-2 (Green band) band-3 (Red band) and band-4
(Near infra red band) were generated. For identifying the villages selected for crop surveys
on the FCC, topographic maps of scale 1:50,000, which contain information on identifiable
features like roads, canals, water bodies etc., were used. Further, a Global Positioning
System (GPS) was used to identify the crop plots on imageries. The GPS was taken to the
plots and location of the plots in terms of longitudes and latitudes were recorded. These
locations were then identified on the FCC’s. The coordinates of each plot in terms of
scanline and column number were also recorded to identify these plots on the Normalized
Difference Vegetation Index (NDVI) imageries. NDVI is defined as the ratio of (IR-R) to
(IR+R) where IR and R denote the radiance in infra red (band 4) and red (band 3)
respectively. The concept of density slicing technique was used to divide the NDVI
IV.214
imageries into three classes named as (i) Non vegetation class, (ii) Average vegetation class
and (iii) the High vegetation class. To obtain the strata weights for the two vegetation
classes, the number of pixels falling into these classes were obtained and since the pixel size
is fixed, hence the area of the classes were obtained which were used as strata weights.
To obtain the post-stratified estimator, let us assume that n villages selected in the sample
have been post-stratified into L strata such that n k villages fall in the k-th post-stratum.
Let Ykij denote the yield for the j-th field in the i-th village of the k-th post stratum.
The sample mean for the k-th post stratum can be defined as
Ykij
yK I J
… (2.1)
mk
where mk is total number of field experiments falling in the k-th post-strata.
Now the post stratified estimator of district average yield can be given by
L/
A y k k
L
y L/
= Wk y k … (2.2)
A k
Ak
Where Ak denotes the area under crop in the k-th post-stratum. Wk
Ak
Ignoring the pre stratification and also ignoring the contribution to sampling error due to post
stratification the Variance and the Estimator of variance of y can be obtained easily given by
The estimates of crop yield based on usual estimator (without using satellite data) and the
post-stratified estimators based on NDVI and RVI are given in table 1 below. From these
results it is seen that the new estimators are significantly more efficient as compared to the
usual estimator of crop yield at district level.
2.3 Small Area Estimation of Crop Yield at Tehsil/ Block Level
Issue of small area estimation has gained importance in view of growing needs of micro level
planning. Most of the small area estimation techniques in the early stages were developed in
the context of demographic studies. Purcell and Kish (1979) categorise these areas under the
general heading of Symptomatic Accounting Techniques (SAT). Gonzalez (1973) described
a small area estimation technique well known as synthetic estimator. In this method an
unbiased estimate is obtained from sample survey for a larger area and this estimate is used to
derive estimates for sub areas having the same characteristics as the larger area.
Let the population (a district) consist of T small areas (Tehsil/Block). Further, let the district
area be divided into V post strata representing crop condition like very good crop, average
crop, poor crop, no crop etc.) based on the vegetation indices derived from the satellite
spectral data. The crop within these post strata is generally homogeneous in respect of the
IV.215
character under study (the crop yield) and the boundaries of these post strata cut across the
small areas. Hence, it can be easily assumed that the units within a small area belonging to
particular post strata will have the same characteristics as the units belonging to those
particular post strata irrespective of the small area.
In order to develop crop yield estimates at tehsil level from general crop yield estimation
surveys based on crop cutting experiments we propose two estimators namely (i) The Direct
estimator and (ii) The Synthetic estimator. These estimators make use of available
information on crop yield and also the information of crop acreage for all the post strata
which overlap the tehsil. It has been seen earlier from the results that stratification based on
NDVI provide more efficient estimate of crop yield for the district as a whole. Therefore, for
small area estimator only NDVI has been used to develop post strata.
2.3.1 Direct Estimator
Let y and x denote the character under study, the crop yield and the auxiliary variable, the
crop acreage respectively. Let y ti denote the crop yield for the ith plot in the tth tehsil and th
post-strata and let ntv denote the number of sample observations belonging to the tth tehsil in
the vth post strata. If all n t ’s are greater than zero then an unbiased post stratified estimator
known as the Direct Estimator for crop yield may be obtained as
y W2 V
V dt tv y tv
... (2.5)
where
2
y s tv
V tv
n tv
n tv 2
2 1
s tv ytvi ytv
n tv 1 i
IV.216
To improve the efficiency of the Direct Estimator, a Synthetic Estimator denoted by y st is
proposed which make use of the information from the whole sample,
X tv
y st w tv y oν , where w t and y o is the average crop yield for the ν th
ν X to
post stratum
IV.217
In case of crop yield modeling using satellite data, several studies have been undertaken to
establish relationship between spectral parameters through vegetation indices and the crop
yield. Sridhar et. al.(1994) presented wheat production forecasting for a predominantly un-
irrigated region in Madhya Pradesh. Singh and Ibrahim (1996) examined the use of multi date
satellite spectral data for crop yield modeling using Markov Chain Model. Saha (1999) used
satellite data and GIS for a developing several crop yield models.
A study on "Evaluation of crop cut method and farmers reports for estimating crop
production"(Verma et al., 1988) was undertaken at Longacre Agricultural Development
Centre UK. This study was carried out in 5 countries in Africa during 1987 with the objective
of comparing crop estimates based on crop cut methods with estimates obtained by asking
farmers directly to state their production. The results of the study showed that farmers eye
estimates are remarkably close to actual production figures in all the countries and they also
show considerably small variance compared to the estimates based on crop cutting
experiments. Therefore considerable interest should be focused on using farmers estimates
which are much cheaper to obtain and easier to conduct.
Singh (2003) made an effort to use the farmers eye estimate more objectively as a auxiliary
variable along with the spectral indices based on remote sensing satellite data to improve the
efficiency of crop yield models for forecasting crop yield. Details of this study are given
below:
3.1 Study Area and Extent of Data used in the Study
The study was conducted for district Rohtak of Haryana State which is one of the major
wheat growing areas having an acreage of more than 66 percent under wheat crop during
Rabi season. In the present study the yield data for the Rabi season for the years 1995-96
and 1997-98 from general crop estimation surveys based on crop cutting experiments for
wheat crop for district Rohtak, Haryana has been used. The satellite data in the study has
been used for 1995-96 from IRS-1B, LISS-II of path 30 and Row 47 of 17th February, 1996.
The total area of Rohtak district is covered in one sub scene B2 of 30-47. For 1997-98 IRS-
1D data of sensor LISS-III of path 95 and row 51 for Feb. 4th, 1998 has been used. A
topographic map is the best tool to supply ground truth information for visual interpretation
and identification of various features on satellite imageries. From these maps locations of
villages along with related features like continuous roads, canals railway tracks etc. can be
easily identified on FCC’s. Survey of India topographical maps of Rohtak district on
1:50,000 scale were used to identify the location of villages selected for the crop yield
estimation surveys.
A Global Positioning System (GPS) was used to identify the locations of the plots selected
for crop cutting experiment for wheat crop in terms of their latitudes/longitudes and also the
locations of ground control points (GCP's) which were later used to rectify the raw digital
spectral data. Farmers yield appraisal data has been collected for the years 1995-96 and 1997-
98 for wheat crop yield from the same farmers for only the fields which have been selected
for crop cutting experiments in general crop estimation surveys.
3.2 Vegetation Indices
Spectral response characteristics of healthy vegetation, can easily be characterised in the
different parts of the electromagnetic spectrum. To further enhance the discrimination
between different spectral vegetation classes, computation of different vegetation indices
using infrared and red band data in the electromagnetic spectrum, for describing the crop
growth conditions, are commonly used. Two most commonly used vegetation indices are:
IV.218
(i) The Normalized Difference Vegetation Index (NDVI) defined as
IR R
NDVI , and
IR R
(ii) The Ratio Vegetation Indices (RVI) defined as
IR
RVI
R
Where IR and R refer to radiance in infrared (band-4) and red (band-3) bands of the satellite.
These two indices have been used in the present study to generate the index images for post-
stratification of the study area on the basis of vegetation vigour.
3.3 Spectral Yield Model
These are empirical models which directly relates the crop yield to the multi-spectral satellite
data or derived parameters in the form of spectral vegetation indices (SVI). In this procedure
SVI at the time of maximum vegetation growth stage of the crop is related to final crop yield
through regression techniques and pre harvest crop yield is forecasted. In India district level
yield models for major crops like wheat, paddy, sorghum etc. have been developed under
crop acreage and production estimation (CAPE) project undertaken by National Remote
Sensing Agency (NRSA), Hyderabad, Deptt. of Space. However these models could explain
about 60% variation in yield and hence are not very efficient.
3.4 Integrated Yield Model using Spectral Data and Farmers Eye Estimate of Crop
Yield
Most of the crop yield models developed so far could not be adopted in practice either
because of delay in the availability of data on different variables to be used in the model or
the high cost in collecting the data and in analyzing the results. For any operational yield
model to be successful for adoption it is necessary that data should be available much before
the harvest of the crop and it should be cost effective. Spectral data in the form of vegetation
indices have proved to be very useful variable for explaining variability of the crop yield
which can be easily available for use in yield forecasting models. Also the farmers eye
estimate of yield is easily available and this information can be used as auxiliary variable
along with the spectral vegetation indices to improve the efficiency of the crop yield models.
An earlier such attempt on using eye appraisal of crop yield of a large number of sample
fields as auxiliary information had been made by Panse, Rajgopalan and Pillai (1966).
In the present study, therefore suitable regression models using spectral vegetation indices in
the form of NDVI and RVI and farmers eye estimate of yield as explanatory variables have
been developed for improved crop yield forecasting models. Both these variables can be
easily obtained at the time of maximum growth of crop.
3.5. Results of the Study Pertaining to Crop Forecasting
The usual linear regression based models have been developed with the crop yield (y) as the
dependent variable and three independent variables, namely RVI (x1), NDVI (x2) and the
farmer’s eye estimate of crop yield of the corresponding plot (x 3). The models have been
developed using the data for the Rabi 1995-96. These models have been used to forecast the
crop yield for Rabi 1997-98 using the independent variables for 1997-98. The results and
predicted value of crop yield using different independent variables independently as well as
together are given in table 2.
IV.219
From this table it is seen that R2 value is 0.45 and 0.54 respectively when only RVI and
NDVI alone are used in the model and it increases to 0.59 when both these variables are used.
However the R2 value is 0.86 when only farmers eye estimate is used as the explanatory
variable and the R2 value increases to around 0.90 when it is used along with RVI or along
with NDVI or along with both RVI and NDVI together. The deviation of the predicted yield
from the actual yield is very low. In almost all the cases it is less than 2%. The standard error
of the predicted value is also small in all the cases and as expected it is minimum when all
the three variables are used together but it is not much different in the case when only farmers
eye estimate and NDVI is used.
The results suggest that a reliable and timely forecast may be obtained using NDVI from the
satellite data along with the farmers eye estimate as the two explanatory variables. Both these
variables can be obtained at the time of maximum vigour of the crop and hence objective
reliable forecast may be made about 6-8 weeks before actual harvest of the crop.
Table 1: Wheat crop yield estimation using stratification based on NDVI and RVI, for
district Rohtak for Rabi 1997-98
IV.220
Table 2: Wheat crop yield forecasting model using RVI (x1), NDVI (x2) and the farmers eye Estimate (x3) as independent
variables for district Rohtak for forecasting crop yield for Rabi 1997-98 . (using the model based on data for Rabi 1995-96).
Model
R2 a b Predicted %S.E. Percentage of
value(Q/hac.) Deviation
ŷ
y a b1 x1 0.451596 3.3445 4.251948 35.86 13.5434 0.1653
(3.102028) (0.5724) (1.998869) (4.8568)
Actual crop yield for Rabi 1997-98 =35.92 (Q/hac) (Figures in braces give the standard error)
IV.221
References
1. Box, G.E.P. and Jenkins, G.M. (1976), Time series analysis: Forecasting and control,
Holden-Day, San Franscisco. 575.
2. Dadhwal, V.K. and Parihar, J.S. (1985), Estimation of 1983-84 wheat acreage of
Karnal district (Haryana) using Landsat MSS digital data, Technical note, IRS-
UP/SAC/CPF/TN/09, Space Application Centre, Ahmedabad.
3. Gonzalez, M.E. (1973) Use and evaluation of synthetic estimates. Proceedings of
American Statistical Association, Social Statistics Section. 33-36.
4. Jain, R.C., Aggarwal, R. and Jha, M.P. (1980). Effects of climatic variables on rice
yields and its forecast. Mausam, 31(4), 591-596.
5. Jha, M.P., Jain, R.C. and Singh, D. (1981). Pre-harvest forecasting of sugarcane
yield. Indian J. Agric. Sci. 51(11), 757-761.
6. Mahalanobis, P.C. (1945). Report on the Bihar crop survey, 1943-44. Sankhya, 7, 29-
118.
7. Panse, V.G., Rajgopalan, M and Pillai, S.S. (1966). Estimation of crop yield for small
domain. Biometrics. 374-384.
8. Purcell, N.J. and Kish, L. (1979) Estimation of small domains. Biometrics. 35, 365-
384.
9. Saha, S.K. (1999). Crop yield modeling using satellite remote sensing and GIS-
current status and future prospects. Processing “Geoinformatics - Beyond 2000, an
international conference on Geoinformatics for natural resources assessment,
monitoring and management” held at IIRS Dehradun during 9-11 March, 1999.
10. Singh, R., Goyal, R.C. Saha, S.K. and Chhikara, R.S. (1992). Use of Satellite Spectral
data in Crop Yield Estimation Surveys. Int. J. Remote Sensing, 13, 2583-2592.
11. Singh, R. and Goyal, R.C. (1993). Use of remote sensing technology in crop yield
estimation surveys. Project Report, IASRI, New Delhi.
12. Singh, R. and Ibrahim, AEI (1996). Use of spectraal data in Markov Chain model for
crop yield forecasting. Jr. Indian Soc. of Remote Sensing, 24 (3), 145-152.
13. Singh, R., Goyal, R.C., Pandey, L.M. and Saha, S.K. (1999). Use of remote sensing
technology in crop yield estimation surveys–II, Project Report IASRI, New Delhi.
14. Singh R, Semwal, D.P., Rai, A. and Chhikara, R.S. (2002). Small area estimation of
crop yield using remote sensing satellite data. Int. Jr. of Remote Sensing. 23(1), 49-
56.
15. Singh, R. (2003). Use of satellite data and farmers eye estimate of yield for crop yield
forecasting, submitted for publication to the JISAS.
16. Sridhar,V.N., Dadhwal, V.K., Choudhary, K.N., Sharma, R., Bairagi, G.D. and
Sharma, A.K. (1994). Wheat production forecasting for a predominantly unirrgated
region in Madhya Pradesh (India), Int. J. Rem. Sens., 15(6), 1307-1316.
17. Sukhatme, P.V. and Panse, V.G. (1951). Crop surveys in India - II. Journal of Indian
Society of Agricultural Statistics 2, 95-168.
IV.222
APPLICATION OF REMOTE SENSING AND GIS IN CROP
ACREAGE ESTIMATION IN NORTH EASTERN HILLY REGIONS
Prachi Misra Sahoo
I.A.S.R.I., New Delhi -110012
1. Introduction
India is predominantly an agrarian economy both from the point of view of employment as
well as contribution to the national income. Availability of reliable and timely agricultural
statistics is hence of paramount importance to the planners, administrators, policy makers and
research scholars. Government depends on these data in taking policy decisions regarding
production, pricing, processing, procurement, storage, transport, marketing, export/import,
public distribution and many other related issues including investment planning. A
continuous evaluation of the mechanism for generation of timely and reliable agricultural
statistics, therefore assumes special significance.
The Directorate of Economics and Statistics (DES) in the Department of Agriculture and
Cooperation (DAC) is the nodal official agency for collection, compilation and publication of
major agricultural statistics like area and production estimates of principal crops. From the
point of view of collection of area statistics, the states in the country are divided into three
broad categories:
States and Union Terrtiores, which has been cadastral surveyed, and where area and land use
statistics of built up are a part of the land records maintained by the revenue agencies
(referred to as “Land Records States” or temporary settled states). The system of land records
is being followed in major states and 4 Union territories of Chandigarh, Delhi, Dadar &
Nagar haveli and Pondicherry. These states/ Union territories accounts for about 86 % of
reporting area.
The states where area statistics are collected on the basis of sample surveys (normally known
as non- land record states or “ Permanently Settled States” which are three in number viz.
Kerala, Orissa and West Bengal). A scheme for Establishment of Agency for Reporting of
Agricultural Statistics (EARAS) has been introduced in these three states which envisages,
inter-alia, either the estimation of areas by complete enumeration or through sample surveys
in a sufficiently large sample of 20% villages/ investigators zone. These states accounts for
about 9% of reporting area.
In the hilly districts of Assam, the rest of the states in North-Eastern Region, Sikkim, Goa,
Union territories of Andaman Islands, Daman & Diu and Lakshwadeep where no reporting
agency had been functioning, the work of collection of Agricultural Statistics is entrusted to
the village headman of the reporting area (5%).
The area statistics are collected on complete enumeration basis in some of the states falling
under category (1) and a scheme for Establishment of Agency for Reporting of Agricultural
Statistics (EARAS) has been introduced in the three states in category (2). While in category
(3) the area statistics are collected on the basis of ad-hoc methods based on impressionistic
approach from the village headman which are purely unscientific and non-statistical. The
estimates of area under different crops are given on the basis of eye estimation only. These
figures are not based on any scientific basis and hence may not be very accurate and reliable.
Therefore, a need to develop a scientific methodology with strong statistical background
which is capable of providing reliable estimates of area under different crops in these regions
was felt.
With the advent of Remote Sensing technology during seventies and its great potential in the
field of agriculture have opened new vistas of improving the agricultural system all over the
world. The remote sensing techniques to obtain the land utilization statistics gained
popularity because of its extensive coverage of geographical area. It is therefore required to
explore the possibility of use of satellite Remote Sensing for collecting agricultural statistics
on a regular basis. The recent advances in the field of geographical techniques, like
Geographic Information System (GIS) increased the potential to change substantially the
statistical approach to study the geographical realities. GIS, is capable of handling
geographical data obtained by different sources like Census, Survey and Remote Sensing.
There is no objective methodology for estimation of area under different crops in North-
Eastern states due to the typical problems existing in these regions. The North-eastern states
particularly Meghalaya, mainly consists of hilly region with thick forest cover. Besides this,
the main problem is its undulating topography and non-accessibility of vast area. Further, the
relative percentage area under the crops is very less. Mostly terraced farming and Jhum
cultivation is practiced in these regions. Moreover, these areas particularly Meghalaya, are
covered by clouds most of the time. Thus it is difficult to get cloud free images of these areas.
Therefore, use of remote sensing satellite data alone may not be able to provide reliable
information. There are no cadastral maps and village boundary maps existing for these
regions. The exact information regarding total number of villages in each district/block is also
not available. Further, within a village’s total number of farmers, number of fields owned by
each farmer, crops grown by the framers etc. is also not known. Thus, the traditional
methodology of area estimation is also not applicable in these regions.
Keeping all this in view, a project entitled “Developing Remote Sensing based methodology
for Collection of Agricultural Statistics in Meghalaya” was taken up at Indian Agricultural
Statistics Research Institute (IASRI), New Delhi in collaboration with Space Applications
Centre (SAC), Ahmedabad and North Eastern Space Applications Centre (NE-SAC),
Shillong. The study was conducted for the state of Meghalaya. Under this project, the
methodology for acreage estimation of paddy crop based on ground survey, remote sensing
and GIS was developed at district level. This article gives the details of the methodology
proposed by Sahoo et al. (2005).
2. Methodology for Estimation of Area under Paddy in Hilly Regions
Two approaches have been followed to obtain the area under paddy in the district.
(1) Village Survey Approach
(2) An integrated approach based on Remote Sensing, GIS along with survey data.
2.1 Village Survey Approach
The village survey approach for estimation of area under crop is followed in those states,
permanently settled such as Kerala, West Bengal and Orissa. Similar approach was adopted
for estimation of area under paddy in the study district. In this approach, 20 villages were
selected through simple random sampling without replacement from the available sampling
frame of the villages in the district. The list of all the farmers growing paddy crop in the
village has been prepared on the basis of information collected from the village headman in
the selected villages. A sample of 5 farmers has been selected from each village from the
sampling frame of paddy growing farmers. The area under paddy in the farmer’s paddy fields
has been recorded by enquiry from the selected farmers. Out of these farmers, 2 farmers were
IV.224
randomly selected for recording area of each of their paddy fields through Global Positioning
System (GPS).
Numbers of problems have been faced during the implementation of this approach. In this
state, system of land records does not exist and the village headman decides its ownership for
a particular period of time. Therefore, the area of the fields is not measured and farmer-
growing crop on a particular field does not have much idea about the area of the field. It has
been found that area of the field reported by the farmer and area of the same field as
measured by GPS has no similarity. This also holds true for the area figures reported by the
village headman. The R2 of linear relationship between actual area of the field as observed by
GPS and area of the field as reported by the farmers is very low i.e., 0.07. Further, there is no
concept of village boundaries existing in the state in a strict sense. Therefore, the exact figure
for total numbers of villages in the district is not known as different agencies are reporting
different figures. Similarly, the nature and size of the villages also differs. The villages,
which are approachable, are of relatively bigger size and more uniform as compared to
villages, which are away from roads. Further, paddy crop is generally grown in the villages,
which are in the valleys or low laying areas. Therefore, estimation of paddy area based on
total number of villages in the district is always likely to provide over estimation of area, as
the list of paddy growing villages are not available. Keeping in view above points, this
approach was not found to be feasible and it is recommended that this should not be adopted
in future.
2.2 Integrated Approach using Remote Sensing and GIS along with Survey data
The traditional methodology of crop area estimation may it be complete enumeration or
sample survey conducted in other parts of the country is not applicable in Meghalaya as no
land record system is existing in the state, there are no cadastral maps and village boundary
maps available. The exact information regarding total number of villages in each
district/block is also not available. Further, within a village total number of farmers, number
of fields owned by each farmer, crops grown by the framers etc. is also not known. Next if
one thinks of applying remote sensing tool for crop area estimation, which is very successful
in the plains, some other problem arises like thick forest cover, undulating topography, non-
accessibility of vast area, terraced farming and Jhum cultivation which makes direct
application of satellite data difficult for crop area estimation. Moreover, these areas
particularly Meghalaya, is covered by clouds most of the time during the year which makes it
difficult to get cloud free images of these areas. Therefore, use of remote sensing satellite
data alone is also not capable of providing reliable information regarding area under different
crops. Keeping all this in view, it was considered that the use of satellite data along with the
ground survey data in GIS environment may be useful to obtain the reliable estimates for the
area under crops. Thus, an integrated approach using remote sensing, extensive ground
survey and GIS is developed to estimate area under paddy crop. The field problems has been
identified and grouped into three major categories and efforts have been made to resolve
these problems in this approach. The three groups of problems identified for crop area
estimation using remote sensing data are the following:
(i) Due to undulating topography of the region, misclassification and angle of the sensor
there may be significant differences of area under crop in the image and actual area under
crop on the ground, which may result in larger extent of misclassification errors.
(ii) The area under paddy crop falling under hill shades or valleys, which may not be
exposed, to the satellite sensor, as satellite sensors are sun-synchronous. Further, small
paddy fields are not detectable due to lower spatial resolution of the LISS-III sensors.
IV.225
(iii) Cloud cover in the satellite image.
The solution for each of these problems is suggested in the study. In order to rectify the area
under paddy crop due to undulating topography and misclassification errors, a relationship
between area under paddy in the classified image and actual area under paddy crop on the
ground has been established. The area under paddy, which has not been captured by satellite
sensor due to hill shades and limitations of spatial resolution of the sensor, has been rectified
by a suitable sample survey in the buffer created along selected roads in GIS environment.
Suitable estimator has been then developed to estimate the area under paddy crop in this
buffer zone. The vector layer of this buffer was overlaid on the classified image and the
corresponding area from the image was extracted. For cloud removal initially composite
technique has been applied. Besides this some other estimators are also developed for
estimating paddy area under clouds or cloud shadow. Using these estimates, an improved
estimate of area under paddy in the entire district was obtained. The detailed steps followed
under this approach are given in the following sections.
2.2.1 Estimation of Area under Paddy by Digital Classification
For estimation of area under a crop choice of the satellite data pertaining to some specific
period in which that crop can be easily identified is very important. In Meghalaya, the winter
paddy period extends from August-September to December-January. Field observation with
satellite imagery shows that in these areas, there may be two critical period for choosing the
satellite data for delineation of paddy areas. During the early vegetative growth period,
because of thin canopy density and visible underneath water gives a typical signature. This
signature helps in delineating paddy area from nearby grassland or forestland, which gives
very bright red signature because of healthy vegetation in the monsoon/post monsoon season.
As per the crop calendar, the September month will be ideal for acquiring satellite image for
winter paddy. The September image for delineation of paddy is advisable as during this
period paddy field was in the puddling and transplanting stage, there by giving a unique
spectral signature which could be easily separated from rest of the neighboring land cover
classes specially other crops grown during that period. If paddy which is cultivated in wet
condition is the only crop grown in this period, any chances of misinterpretation because of
other crop grown in wet condition are also eliminated. Another time period in which crop will
be easily identified is when it will be in mature stage. During this time, image will give
typical signature, which will allow delineating paddy crop from nearby vegetation. Images
acquiring during the month of November will be ideal for this purpose. It is observed that in
the month of December and January the areas with barren hills vegetation cover used to dry
up and those signature get mixed with harvested paddy fields. Thus, in this period it will be
difficult to delineate the paddy area from nearby neighboring vegetation cover. Maximum
likelihood classifier was found to be accurate enough for classifying the study area for
extraction of paddy fields.
2.2.2 Estimation of Area Under Cloud
Cloud cover is a very complex problem in optical remote sensing especially over the humid
tropical regions. One of the most common solutions to this problem is to produce a cloud free
mosaic from several multi-date images acquired over the same area of interest. In this method
the image containing the least cloud cover is taken as the base image. The cloudy areas in the
image are masked out, and then filled in by the cloud free areas from other images acquired at
different time. This is known as compositing technique for cloud removal. The cloud in the
images were removed using estimation of area under cloud using compositing technique.
Under this approach, data for two different years has been classified. One is the current year
data containing clouds and another is the data for past year, which is cloud free. Then the area
IV.226
or patches, which are containing clouds/ cloud shadow in the current year image are
substituted with the patches obtained from cloud free image of past year. We may take the
data of previous year, as the past year data as generally there is not much change in the area
under crops from year to year.
2.2.3 Development of Relationship between Area under Paddy Crop in the Image and
Actual Area under Paddy on the Ground
To develop a relationship between area under paddy in the image and actual area under paddy
on the ground a suitable number of paddy patches/ fields, which are clearly visible and de-
markable in the False Colour Composite (FCC) of the study area have been identified. The
actual area of these patches / fields was recorded on the ground by using GPS. Further, these
patches were extracted from the classified digital image and the corresponding area under
paddy was recorded.
Let there be ‘nr’ clearly visible paddy patches/ fields, which can be demarcated in the FCC as
well as on the ground. Let Y(r)i and X(r)i denote the area under paddy crop as obtained by GPS
on the ground and from the classified image for ith paddy patch respectively. The model can
represent the linear relationship between area measured by GPS and obtained by classified
image:
Yr i αr βr Xr i er i … (1)
IV.227
2.2.4.2 Development of a Linear Relationship between Area under Paddy Through Eye
Estimate and actual Measurement taken through GPS
Let Y(be)hji denotes the area under paddy in ith grid of jth road from hth category as recorded by
eye estimate. Similarly, Y(bg)hji denotes the area under paddy crop in ith grid of jth road from hth
category as measured by GPS; where i=1, 2, 3,.. m(bg)hj and m(bg)hj m(b)hj as the area under
paddy has been recorded only for approachable grids.
The area under paddy through eye estimate from those segments, where it was not measured
by GPS was corrected by following equation.
ŷbghji = α̂(b) β̂ ybehjib … (2)
M L N b h m b hj
Ŷ y b hj
… (3)
m
b bg hji
h 1 j 1 i 1
b hj
Ŷ(b)
ŶP Ŷ(r) … (4)
Ŷ(br)
3. Conclusions
The methodology developed for the estimation of area under paddy crop is widely accepted
and is extended for other districts of the state. It has become operational for the entire state of
Meghalaya. The methodology will also be extended to other North-eastern states like
Manipur and Mizoram. The methodology for the acreage estimation for multiple crops is still
in development stage and research is going on in this direction.
IV.228
References
1. Sahoo, P.M., Rai, A., Singh, R., Handique, B.K. and Oza, M.P. (2007). Development
of Remote Sensing based Methodology for collection of Agricultural Statistics in
Megahalaya, Project Report. IASRI New Delhi, Publication.
2. Sahoo, P.M., Rai, A., Krishnamoorthy, S., Handique, B.K., Rao, P.P. N., Oza, M. P.
and Parihar, J.S. (2006). Sampling Approach for Estimation of Crop Acreage under
Cloud Cover Satellite Data in Hilly Regions. Presented at 5th Asia-Pacific Remote
Sensing Symposium held by International Society of Optical Engineering (SPIE) in
Goa from Nov. 13-17.
3. Sahoo, P.M., Rai, A., Singh, R., Handique, B.K., Oza, M. and Parihar, J.S. (2006).
Remote Sensing and GIS based Sampling Methodology for Estimation of Crop Acreage
in North-Eastern Hilly Region. Presented in 9th Annual International conference and
Exhibition in the field of geographic information, science, technology and application
Map India 2006 held at New Delhi from Jan 30 – Feb 1, 2006.
4. Sahoo, P.M., Rai, A., Singh, R., Handique, B.K., and Rao, C.S. (2005). Integrated
Approach Based on Remote Sensing and GIS for Estimation of Area under Paddy Crop
in North-Eastern Hilly Region. Journal of Indian Society of Agricultural Statistics,
59(2), 151-160.
5. Sahoo, P.M., Rai, A., Singh, R., Handique, B.K., Oza, M. and Parihar, J.S. (2005).
Remote Sensing based sampling methodology for estimation of Crop Acreage in North-
Eastern Hilly Region in the 25th ISRS Annual Convention and National Symposium
on Emergence of Geoinformatics for Development: trends & opportunities, held from
December 6-9, 2005 at Birla Institute of Technology, MESRA, Ranchi.
6. Sahoo, P.M., Rai, A., Singh, R., Handique, B.K., Oza, M. and Parihar, J.S. (2005).
Remote Sensing and GIS based approach for estimation of Crop Acreage in North-
Eastern Hilly Region in the Workshop on “Improvement of Agricultural Statistics” held
during 01-02 July 2005 at Vigyan Bhavan, New Delhi by Ministry of Agriculture
IV.229
REMOTE SENSING FOR CROP GROWTH AND
CROP SIMULATION MODELING
Vinay Sehgal
I.A.R.I., New Delhi -110012
1. Introduction
Crop yield is the result of complex interaction among factors of soil, atmosphere, plant
genotype, and management practices adopted. These complex interactions of crop with
various factors and of factors among themselves make the crop yield modeling a difficult
task. Several methods of crop yield forecasting have been developed ranging from purely
statistical to agro-meteorological to empirical biophysical to mechanistic simulation models
varying in complexities and input data requirements. Based on the methodology, all these
yield models can be divided into two categories: (a)Empirical/Regression and (b) Simulation.
The regression models assume that the variability in crop yield can be explained only by a
few independent variables considered which are independent of each other. These regression
models have many shortcomings due to errors in these assumptions. In contrast, simulation
models closely represent a real crop by gradually growing leaves, stems, roots etc in response
to external influences. In these models various mechanisms of photosynthesis, respiration,
dry matter production and crop phenology are dynamically linked to each other by various
rate equations and each mechanism is defined by the value of its state variables at the end of
simulation time scale of one day. Some more complex simulation models include the
subroutines of soil water and nitrogen balance to account for the effects of water and nitrogen
stress on dry matter production. These models have good yield predictive capacity but are
input data intensive. Dynamic growth simulation models are reported to have universal
applicability compared with statistical models which are largely site-specific (Jamieson et al.,
1991).
Though the crop simulation models are very useful in evaluating crop growth at field scale,
their implementation at larger scale is restricted by the input data availability at
corresponding scale. Chipnashi et al. (1999) observed that most crop models are applicable at
individual plot scale, a unit of analysis that is too small for meaningful assessment of weather
impacts as experienced at farm, district or provincial level. The remote sensing system
directly views the above-ground component of crop canopy over large areas with possibility
of deriving parameters for each field. A combination of remotely sensed information and crop
models can provide a more direct link between canopy state variables and remotely sensed
data, and is more independent on the time scale that the remotely sensed data is acquired. A
Geographic Information System (GIS) provides the ideal integration environment for spatial
remotely sensed data, other spatial and aspatial data and crop simulation model.
2. Coupling Satellite Observations and Crop Simulation Models
Crop simulation models, which are used to monitor crop growth and estimate yields, have
limitations to their application at regional scales. The limitations, at this scale, are linked to
the difficulty of estimating some of the model input parameters and/or model initial
conditions (Moulin and Guerief, 1999). Besides, models may not produce results of desired
accuracy for practical applications due to (i) over-simplification of some of the processes
modelled and (ii) non-incorporation of a number of biotic (e.g. pest, diseases) and a-biotic
(e.g. micro-nutrients, salinity) stress factors which operate on farmer’s fields. Moulin et al.
(1998) has suggested that combining crop model and remotely sensed information is a
promising approach to overcome some of the mentioned limitations especially at regional
scales.
The basis of this RS-model coupling is that remotely sensed measurements can be related to
the instantaneous values of various canopy state variables. To achieve this, canopy state
variables can be connected to remotely sensed measurements either by using physical
radiative transfer models or empirical relationships. The main advantage of using remotely
sensed information is that it provides a quantification of the actual state of crop (for large
area) using methods that are less labour and material intensive than in situ sampling. While
crop models provide a continuous estimate of growth over time, remote sensing provides a
multispectral assessment of instantaneous crop condition with in a given area (Delecolle et
al., 1992). The added value of crop models and remote sensing is that they are objective,
quantitative and consistent over large areas.
The different ways to combine a crop model with remote sensing observations (radiometric
or satellite data) were initially described by Mass (1988a) and his classification scheme with
examples was revisited by Delecolle et al. (1992) and by Moulin et al. (1998). Five methods
of data integration have been identifies:
(i) the direct use of a driving variable estimated from remote sensing information in
the model;
(ii) the updating of a state variable of the model (e.g. LAI) derived from remote
sensing, also called ‘forcing’ strategy;
(iii) the re-initialization of the model i.e. the adjustment of an initial condition to obtain
a simulation in agreement with the remotely-sensed derived observations;
(iv) the re-calibration of the model i.e. the adjustment of model parameters to obtain a
simulation in agreement with the remotely-sensed derived observations, also called
‘re-parameterization’ strategy;
(v) the corrective method i.e. a relationship is developed between error in some
intermediate variable as estimated from remotely sensed measurement and error in
final yield. This relationship may be applied to a case in which final yield is not
known.
The forcing strategy consists of updating at least one state variable in the model using remote
sensing data. The re-initialization method takes advantage of the dependence of model
performance on state variable initial condition. It involves adjustment of initial condition of
state variable so as to minimize the difference between a derived state variable or the
radiometric signal and its simulation. Re-calibration method considers remotely sensed
information as an absolute reference for the model simulation. The basic assumption is that
the model, although formally adequate, may be inaccurately calibrated. So, this method
consists of adjusting model parameters in such a way so as to minimize error between remote
sensing derived state variable and its simulation. In corrective approach, an empirical
relationship is established between error in the state variable and error in the final yield. For
this approach, few known sites are selected where the state variable is measured through
remote sensing as well as the final yield values are known. So, the error in simulation of state
variable is computed along with the error in the final yield. An empirical relationship is
developed between these two errors. Now, for unknown sites, the actual value of state
variable is estimated from remote sensing measurements and from it the error in simulated
state variable is estimated. Then using this error in the empirical relationship, yield of
unknown sites is estimated.
IV.231
3. Coupling RS, GIS and CSM: Crop Growth Monitoring System (CGMS)
A “Crop Growth Monitoring System” (CGMS), linked to a GIS, assimilates crop
environment information from various sources spatially and passes them to a crop simulation
model for simulating crop growth and yield at regional scale. The CGMS permits spatial
analysis of crop growth by running crop simulation model at a regular geographical grid. The
grid-wise simulation model inputs are generated from spatial layers of weather variables, soil
properties, crop distribution, cultivar parameters and crop management practices in a GIS
environment A CGMS based on WOFOST model is used for operational yield forecasting of
important crops grown in 12 member states of European Union (Meyer-Roux and Vossen,
1994). A CGMS based on EPIC model has been demonstrated by Priya et al. (1998) working
at two different grid resolutions of 50 km and 10 km for predicting maize, wheat and rice
yields in India at national and regional scales.
A framework of implementing the wheat crop simulation model WTGROWS for regional
yield prediction using remote sensing information and required spatial and non-spatial
databases in a GIS environment is presented here. The issues related to framework design,
generating required inputs, and its implementation as a crop growth monitoring system
(CGMS) has been discussed. The performance of the CGMS has also been evaluated in terms
of district level yield prediction for two rabi seasons of 1996-97 and 2000-01 for the State of
Haryana.
3.1 CGMS Prototype
The broad framework of crop growth monitoring system showing inter-linkages between
geo-informatics tools and crop simulation model and the flow of data is shown in figure 1. It
consists of four major components, namely, (a) inputs assimilated in GIS, (b) a relation
database management system (RDBMS), (c) a two-way linking shell between RDBMS and
crop model and (d) crop simulation model WTGROWS. The framework has been
implemented on MS Windows XP™ platform on a personal computer. The MS ACCESS™
software has been used as RDBMS while ENVI™ AGROMA™ softwares have been used
for image processing and GIS functions. All the spatial layers for the study area were
georeferenced in UTM (Universal Transverse Mercator) projection Zone 43 North with
Indian Datum.
Grid Layer: A 5’x5’ polygon vector grid layer was generated for the state of Haryana in GIS.
Each grid cell represents one simulation model run and hence all the other inputs were
assimilated/aggregated at grid cell level in the RDBMS though they were having information
at different spatial scales. The serial number i.e. identifier, area and central latitude of each
grid cell was generated and stored in RDBMS.
IV.232
Figure 1 Schematic diagram of a crop growth monitoring system showing the linkages between inputs, spatial layers in GIS,
and relational database to WTGROWS simulation model.
GN+FRACTION CROP
RS DATA CLASSIFIED CROP MASK
AREA
IV.233
PTF GN + SOIL WATER
SOIL TEXTURE CLASS
CONSTANTS
SOIL MAP (NBSS&LUP)
SOIL DEPTH CLASS GN + SOIL DEPTH
CGMS SHELL ver 1.0
SOIL OC SURFACE
GN + SOIL OC %
SOIL OC% (POINTS)
(INTERPOLATION)
DISTRICT-WISE CROP
GN + CULTIVAR,
MANAGEMENT
SOWING DATE,
Cultivar, Sowing Date,
IRRIGATION, N FERTI
Irrigation, Fertilizer
DISTRICT-WISE
PTF – Pedo Transfer Functions AGGREGATED
CROP YIELD
Selection by SHELL as per run parameters
3.1.1 Administrative Boundary Layer
Vector layer of district boundaries were digitized from 1:250,000 scale Survey of India (SOI)
maps. The district boundary layer was overlaid on grid layer and each grid was assigned a
district code depending on the maximum district area in the grid.
3.1.2 Soil Properties Layers
The soil resource map of Haryana produced by NBSS&LUP (Sachdev et al., 1995) at
1:250,000 scale was digitized. Soil depth and soil texture layers were generated after
reclassifying the soil-mapping units according their attributes of depth and particle size,
respectively. Each grid cell was assigned average soil depth and dominant soil texture class.
To account for soil fertility, soil organic carbon raster map was produced by interpolating
from 72 point data collected from literature using inverse square distance interpolation. An
average soil organic carbon content was calculated for each grid cell and stored in RDBMS.
3.1.3 Weather Database and Surfaces
The daily weather data (Rainfall, Temperature maximum and minimum, Wind speed, and
Relative Humidity) of 17 surface observatories in and around Haryana State were entered as
table in RDBMS for the crop year 1996-97. A weather data interpolation program was
written in VISUAL BASIC™. The program read daily weather data of observatories with
their locations from the database tables and generated daily surface of each weather
parameter at 5’X5’ resolution using inverse square distance interpolation. Boring through the
daily weather surfaces generated grid-wise daily weather data file in the format of
WTGROWS.
Due to the non-availability of daily solar radiation for most of the observatories, Hergreave’s
method of estimating daily solar radiation from temperature range was adopted. The
coefficients of Hergreave’s equation for various stations in wheat belt of India have been
derived. Using the geographical coordinates of such stations in and around Haryana State, a
thiessen polygon surface was generated and each grid cell was assigned dominant polygon’s
Hergreave’s coefficients.
3.1.4 Crop Simulation Model
The WTGROWS (Aggarwal et al., 1994) is a detailed production level 3 mechanistic model,
which simulate the potential production, phenology, soil water balance, soil and plant
nitrogen balance and water and nitrogen stress on plant growth and development. It has
limitation that it does not simulate the effect of biotic stresses (pests and diseases) on crop
growth and development. It requires inputs on site data, daily weather data, soil
characteristics and crop management data. The model has been well calibrated for Indian
wheat cultivars. In this study, the standard values of genetic constants for a semi-dwarf
medium duration high yielding wheat cultivar given by (Aggarwal et al., 1994) were adopted.
The model, written in PCSMP (Personal Computer Continuous System Modeling Program by
IBM, 1975), runs on any IBM compatible PC under MS-DOS.
3.1.5 Crop Distribution Layer
For each of the two seasons, multi-date IRS-WiFS images were acquired, registered and geo-
referenced. Using hierarchical decision rule technique, multi-date images were classified and
wheat pixel mask was generated showing wheat distribution in the study area. Fraction of
wheat area to grid cell area was calculated for each grid and stored in grid attribute table in
RDBMS.
IV.234
3.1.6 Crop Management Inputs
The three important management inputs of sowing date, amount and dates of irrigation and
type, rate and dates of Nitrogen fertilizer application were specified district-wise, based on
literature and expert judgment, as table in RDBMS. This table was joined to grid attribute
table on the common field of district code so that the management inputs could be specified
for each grid cell.
3.1.7 CGMS Interface Shell
For interfacing the spatial inputs generated as grid attribute table to WTGROWS model, the
“linking” strategy described (Hartkamp et al., 1999) was adopted. The linking shell was
written in C language, which read the grid attribute table and generated the required input
parameters for the model for each grid having wheat area. It also copied the daily weather file
for each grid as the current weather file.
Pedo-transfer functions were incorporated into the shell to generate volumetric soil-water
constants from the grid cell textural class. The pedo-transfer function coefficients were
generated from the experimental soil dataset of twenty locations in Haryana published by
(Komos et al., 1979). The organic nitrogen in soil was initialized for each grid cell from
organic carbon content by assuming a C:N ratio of 10:1. The shell also initialized soil
moisture at sowing as 75 percent of field capacity.
The shell ran the model for each of the grid and model outputs were written back into the grid
attribute table in the RDBMS to be visualized as grain yield and biomass maps in GIS. The
error trapping was also built into the shell to know if model simulation could not be
accomplished for any of the grid cell.
The district-wise average wheat yields were computed in the RDBMS from weighted average
yield values of grids belonging to a district using wheat area to grid area fraction as weight.
The CGMS performance was evaluated by comparing its prediction of district level wheat
yield with the observed yield for each of the two crop years. District level observed wheat
yields were obtained from the State Department of Agriculture of Haryana.
3.2 Results of CGMS
Haryana was covered in 708 grid cells of 5’X5’ and the area of grid cells varied between
73.54 to 75.94 km2. The soil texture classes were dominated by loamy class (338 grid cells),
only two and twenty grid cells had fine and skeletal loamy texture class. The remaining grid
cells were distributed between sandy (192 cells) and coarse loamy (156 cells) classes. In the
case of average organic carbon, 493 grid cells had values between 0.2 – 0.4 percent, 162
between 0.4 – 0.6 percent while 24 grid cells had above 0.6 and 29 below 0.2 percent. The
classified image generated from multi-date WiFS data for 1996-97 indicated that 18 percent
of grid cells had wheat area fraction of less than 0.25, 14 percent between 0.25 – 0.50, 30
percent between 0.50 – 0.75, 37 percent above 0.75. For 2000-01, the multi-date WiFS
classified image indicated that 35 percent of grid cells had wheat area fraction of less than
0.25, 21 percent between 0.25 – 0.50, 24 percent between 0.50 – 0.75, and 20 percent above
0.75.
The output parameters extracted from the grid-wise simulation in addition to the final grain
yield were pre and post anthesis durations, leaf area index (LAI) at anthesis and total above
ground dry matter (TDM) at harvest. The range and arithmetic means over the wheat grid
cells of Haryana for 1996-97 and 2000-01 are summarized in Table 1. The pre-anthesis
durations differed by 15 days (1996-97) and 14 days (2000-01), a reduction from the
specified 20-day variation in sowing dates amongst the districts. The range in post-anthesis
IV.235
durations was further reduced to only 4 and 3 days for 1996-97 and 2000-01, respectively.
Thus, WTGROWS model was able to capture the observed behavior of narrow range in
maturity dates in spite of large variation in sowing dates. The LAI at anthesis and TDM also
showed wide variations among grids. The results prove the capability of CGMS framework in
capturing the spatial variability in wheat phenology and growth as a result of spatial
variations in different inputs. The CGMS framework produced daily spatial distribution maps
of each of these growth parameters that are not shown here.
Table 1: Range and mean of grid-wise simulated wheat growth and yield characteristics for
Haryana for two study years.
The final grain yields varied from 3.5 to 4.75 t ha-1 with a mean value of 4.08 t ha-1 for 1996-
97. For the year 2000-01, the final grain yield varied between 3.29 to 4.73 t ha-1 with a mean
value of 3.94 t ha-1. The grid-wise simulated grain yields are shown in figure 2(a) and 2(b) for
years 1996-97 and 2000-01, respectively. The figures clearly indicate spatial patterns in yield
variability as captured by CGMS. The high grain yields in Kurukshetra and Karnal and low
yields in Bhiwani, Rohtak, and Gurgaon districts are brought out clearly in both the years.
Broadly, the patterns of yield variations are similar in both the years though many grids have
significantly different yield values, such as in Yamunanagar and Rewari districts. This may
be attributed to the differences in weather conditions prevailing in the two years.
While the grain yield cannot be validated at grid level with published ground based estimates,
they can be compared at district level aggregation. This test was carried out to evaluate the
overall performance of CGMS framework. The IRS WiFS derived wheat area estimate of
each grid cell was used as weight in aggregating grid values at district level. The comparison
of simulated grain yields aggregated at district level and estimates by the State Department of
Agriculture is shown in figure 3(a) and 3(b) for years 1996-97 and 2000-01, respectively.
For the year 1996-97, in general, the model simulated yields were higher than observed with
a correlation coefficient of 0.63. The model predicted yields were within 10% of reported
yields in 12 out 16 districts. A RMSE of 335.4 kg ha-1, which is less than 10 percent of the
State mean yield, was obtained.
For the year 2000-01, a correlation coefficient of 0.74 was obtained between predicted and
observed yields. The model underestimated and overestimated yield values for approximately
equal number of districts. The districts with low observed yield values were overestimated
while districts with high observed yield values were underestimated. The model predicted
IV.236
yields were within 10% of reported yields in 14 out 16 districts. A RMSE of 314.1 kg ha-1
was obtained which is 8% of State mean yield.
The model validation results for the two years show that CGMS effectively picked up the
variations in yield among districts with an average error of less than 10%. There is a close
matching between observed and predicted yield values.
Figure 2. The wheat yield map generated by CGMS framework for Haryana, (a) 1996-
97 season, and (b) 2000-01 season.
4000 4000
3500 3500
r = 0.63 r = 0.74
RMSE = 335.4 RMSE = 314.1
3000 3000
3000 3500 4000 4500 5000 3000 3500 4000 4500 5000
Observed Yield (kg/ha) Observed Yield (kg/ha)
Figure 3. Comparison of observed and simulated district-wise wheat yields under for
Haryana (a) 1996-97 season, and (b) 2000-01 season.
IV.237
For the year 1996-97, in general, the model simulated yields were higher than observed with
a correlation coefficient of 0.63. The model predicted yields were within 10% of reported
yields in 12 out 16 districts. A RMSE of 335.4 kg ha-1, which is less than 10 percent of the
State mean yield, was obtained.
For the year 2000-01, a correlation coefficient of 0.74 was obtained between predicted and
observed yields. The model underestimated and overestimated yield values for approximately
equal number of districts. The districts with low observed yield values were overestimated
while districts with high observed yield values were underestimated. The model predicted
yields were within 10% of reported yields in 14 out 16 districts. A RMSE of 314.1 kg ha -1
was obtained which is 8% of State mean yield.
The model validation results for the two years show that CGMS effectively picked up the
variations in yield among districts with an average error of less than 10%. There is a close
matching between observed and predicted yield values.
4. Conclusion
Production forecasting of major cereal crops is a vital component of national food security
and is of immense value to planners and decision-makers both in government and private
sectors. Crop production forecasting is a challenging task because the crop growth and
development is influenced by a large number of environmental factors resulting in large year-
to year and location-to-location variations in crop yield.
Dynamic crop simulation models are one of the tools which have been applied for crop yield
estimation / forecasting. Though these models have good yield predictive ability, their
usefulness has been restricted to applicability at research plot scale due to high input data
requirements. Their implementation at the scales of farmers’ fields, district and province is
restricted by the input data availability at corresponding scales. Also at regional scale,
simulation model may not produce results of desired accuracy due to the specification of
average input conditions, errors in model initial conditions and model parameter
specification.
A combination of remote sensing information and crop simulation models provides ideal
strategy for regional level crop yield estimation. Remotely sensed information can either be
used directly as input to crop simulation model or can be used for in-season model course
correction by adjusting model parameters or initial conditions.
In order to flag the issues related to spatial implementation of simulation model for regional
scale, a case study of development of a prototype CGMS based on crop simulation model and
geo-informatics tools of GIS, RDBMS and remote sensing has been presented. The CGMS
could assimilate spatial inputs of weather, soil, crop management, RS-derived crop
distribution and link it with wheat simulation model WTGROWS to generate crop growth
parameters and yield maps on a daily time step. As simulation models use daily inputs and
compute the daily status of crop growth, this tool is fully capable of continuous in-season
crop assessment. The simulation model was able to capture the variability in wheat yield
within and across districts and the grid-wise weights derived from wheat distribution map
allowed for estimation of average district level yields with an average error of less than 10%.
In the proposed CGMS framework, remote sensing data has been used only for generating
crop distribution map and using it for aggregating grid-wise results to district level. But this
framework can be adapted to assimilate remote sensing derived crop phenology and crop
biophysical products, such as LAI and FAPAR from MODIS sensor, for input to crop
simulation model. Besides, the crop biophysical products can be used for mid-season course
correction of simulation model which may compensate for some of the errors in model inputs
IV.238
and initial conditions, especially at regional scale. These uses of remote sensing information
will result in better accuracy in capturing spatial variability in crop growth and yield across
the region.
References
1. Aggarwal, P.K., Kalra, N., Singh, A.K., and Sinha, S.K. (1994). Analyzing the
limitations set by climatic factors, genotype, water and nitrogen availability on
productivity of wheat I. The model description, parameterization and validation, Field
Crops Research., 38, 73-91.
2. Chipnashi, A.C., Ripley, E.A., and Lawford, R.G. (1999). Large-scale simulation of
wheat yields in a semi-arid environment using crop-growth model, Agric. Sys., 59, 57-
66.
3. Delecolle, R., Maas, S.J., Guerif., M. and Baret, F.(1992). Remote sensing and crop
production models: present trends, ISPRS J. Photogramm. Remote Sens., 47, 145-161.
4. Hartkamp, D.A., White, J.W., and Hoogenboom, G. (1999). Interfacing geographic
information systems with agronomic modeling: A review, Agronomy Journal, 91, 761-
772.
5. Jamieson, P.D., Porter, J.R. and Wilson, D.R. (1991). A test of the computer simulation
model ARCWHEAT1 on wheat crops grown in New Zealand, Field Crops Res., 27,
337-350.
6. Komos, F., Oswal, M.C. and Khanna, S.S. (1979). Soil water functional relationships of
some soils of Haryana State, Journal Indian Society of Soil Science, 27(4), 345-354.
7. Maas, S.J. (1988a). Use of remotely-sensed information in agricultural crop growth
models, Ecol. Modelling, 41, 247-268.
8. Meyer-Roux, J. and Vossen, P. (1994). The first phase of the MARS project, 1988-
1993: overview, methods and results, Proc. of the Conf. on the MARS Project:
Overview and Perspectives, Commission of the European Communities, Luxembourg,
33-85.
9. Moulin, S., Bondeau, A. and Delecolle, R. (1998). Combining agricultural crop models
and satellite observations: from field to regional scales, Int. J. Remote Sens., 19(6),
1021-1036.
10. Moulin, S. and Guerif, M. (1999). Impacts of model parameter uncertainties on crop
reflectance estimates: a regional case study on wheat, Int. J. Remote Sens., 20(1), 213-
218.
11. Priya, S., Shibasaki, R., and Ochi, S. (1998). Modelling spatial crop production: A GIS
approach, Proc. 19th Asian Conference on Remote Sensing, Manila, A-9-1 – A-9-6.
12. Sachdev, C.B., Lal, T., Rana, K.P.C. and Sehgal, J. (1995). Soils of Haryana for
optimizing land use, NBSS Publ. 44 (Soils of India Series), National Bureau of Soil
Survey and Land Use Planning, Nagpur, India, 59 p + 2 sheets soil map (1:500,000
scale).
IV.239
ESTIMATION OF AREA UNDER AGROFORESTRY USING
REMOTE SENSING
Tauqueer Ahmad
I.A.S.R.I., New Delhi -110012
1. Introduction
One of the most important problems faced by the developing countries is producing food in
adequate quantities and of good quality for the fast growing population. It is also necessary to
ensure that the quality of land resources that are directly or indirectly utilized in producing food
is maintained and improved. Therefore, it is important to increase productivity and at the same
time to conserve and enhance the quality of eco-system.
Agroforestry is a land management and farming system that are not only capable of producing
food from marginal agricultural land but also capable of maintaining and improving the quality
of environment. The integration of farming with forestry practices on the farm for the benefit of
agriculture is known as agroforesty. Agroforestry plays a vital role in achieving integrated rural
and urban development. The National Agriculture Policy (2000) emphasized the role of
agroforestry for efficient nutrient cycling, nitrogen fixation, organic matter addition and for
improving drainage and underlining the need for diversification by promoting integrated and
holistic development of rainfed areas on watershed basis through involvement of community to
augment biomass production through agroforestry and farm forestry. The Task Force on
Greening India for Livelihood Security and Sustainable Development of Planning Commission
(2001) has also recommended that for sustainable agriculture, agroforestry may be introduced
over an area of 14 million ha out of 46 million ha irrigated areas that are degrading due to soil
erosion, water logging and salinization. For integrated and holistic development of rainfed areas,
agroforestry is to be practiced over an area of 14 million ha out of 96 million ha. This all will,
besides ensuring ecological and economic development provides livelihood support to about 350
million people. The practice of agroforestry can help in achieving these targets. Therefore, in the
quest of optimizing productivity, the multi tier system came into existence. Gap of demand and
supply of forest produce in India is widening and forests are unable to fulfill the demand.
Agroforestry can play an important role in filling this gap and conservation of natural resources.
The origin of agroforestry practices is believed to have been during Vedic era (Ancient period,
1000 BC) but the agroforsetry as science is introduced only recently. The systematic research in
agroforestry geared up after the establishment of the International Council for Research in
Agroforestry (ICRAF) in 1977, which was renamed in 1991 as International Centre for Research
in Agroforestry. During 2001-02, ICRAF adopted a new brand name “World Agroforestry
Centre”, to reflect more fully their (ICRAF’s) global reach and also their more balanced research
and development agenda; however their legal name “International Centre for Research in
Agroforestry” will remain unchanged.
In India, organized research in agroforestry was initiated in 1983 by the establishment of All
India Coordinated Research Project (AICRP) on Agroforestry by Indian Council of Agricultural
Research (ICAR) at 20 centres and later establishment of the National Research Centre for
Agroforestry at Jhansi in 1988. At present 39 centres of agroforestry are working in the county.
But there is no reliable data available for area under agroforestry in the country and there is no
scientific methodology available for generation of area statistics under agroforestry. Therefore, it
was proposed to develop a methodology for estimation of area under agroforestry using Remote
sensing and GIS techniques under the study entitled “National Initiative on Climate Resilient
Agriculture-Agroforestry Component (NICRA-AF)”. Remote sensing and GIS has emerged as
powerful tool for planning and decision support in the area of agricultural research and
mangement. Estimation of crop area statistics is one of the important field in which this
technology has been used very successfully. Estimation of forest cover is being done by Forest
Survey of India (FSI) using Remote Sensing techniques but for estimation of tree cover,
sampling methodology is being used by FSI. However, remote sensing satellite data is being
used for stratification in estimation of tree cover. Therefore, it was proposed to estimate area
under agroforestry using high resolution satellite imageries (LISS IV) under this study.
2. Study Area
For development of the proposed methodology Ludhiana district of Punjab State was selected on
pilot basis. Ludhiana district is one of the twenty districts of Punjab. The district has an area of
3685 sq. km. with a population of 30,30,352 (2001 Census). Its head quarter is at Ludhiana,
which is located in central part of the district and is at far away distance from the State Capital,
Chandigarh. There are twelve (12) blocks & seven (7 tehsils) and the total number of villages is
915. The density of population per square kilometer is 804 (Source: NIC, Ludhiana, Census-
2001).
Ludhiana district has an area of 3,685 sq. km. and it is the fourth largest district of Punjab. It is
the most centrally located district which falls in the Malwa region of Punjab State. It lies
between 300 34' – 310 01' N Latitude and 750 18' – 760 20' E Longitude. It is bounded on the
north by Sutlej river which separates Ludhiana from Jalandhar district. It shares common
boundaries with Rupnagar and Fatehgarh districts in the East and Moga and Firozpur districts in
the West, Barnala district and Sangrur districts in the South respectively. Ludhiana district is
well developed in transport and communication facilities. The National Highway No.1
originating from Amritsar to National Capital New Delhi passes through the district.
3. Data used in the Study
The data used in the study area is given below:
(i) Satellite Data
The satellite data used for the study for Ludhiana district of Punjab State is mentioned in Table 1.
IV.241
8. 13-Oct-2009 202/027 LISS-IV IRS-P6 NRSC
9. 20-Mar-2010 202/041 LISS-IV IRS-P6 NRSC
10. 27-Oct-2009 202/041 LISS-IV IRS-P6 NRSC
11. 27-Oct-2009 202/042 LISS-IV IRS-P6 NRSC
12. 27-Oct-2009 202/043 LISS-IV IRS-P6 NRSC
The above mentioned satellite data purchased from National Remote Sensing Centre(NRSC),
Hyderabad was in Geo-TIFF format. The data being LISS IV, resolution is 5.8 metre and is in
three bands (Green: 0.52 to 0.59µm, Red: 0.62 to 0.68µm, NIR: 0.77 to 0.86µm).
(ii) Collateral Data
Besides satellite data other data used for the study are:
1. List of villages obtained from Survey of India (SOI), Dehradun
IV.242
(iii) Edge Matching
Edge matching is a procedure to adjust the position of features extending across map sheet
boundaries. This function ensures that all features that cross adjacent map sheets have the same
edge locations. In edge matching technique, matching of similar feature class from two geo-
referenced satellite images is done. Here, edge matching was done using Auto-sync tool of
ERDAS Imagine software.
(iv) Mosaicing of Images
The georeferenced images were joined together and a larger image was formed.
(v) Subset of Area of Interest (AOI)
It is a process of delineation of area of interest (region of interest) from the large image.
Therefore, AOI was delineated for the district under study.
(vi) Ground Truthing
Extensive ground truthing for image analysis was done in 20 villages of 7 blocks of Ludhiana
district in the month of November 2011 and in 22 villages of 8 blocks of Vaishali district in
December 2011. The data collected through GPS during ground truthing was treated as training
site data and was applied to whole image.
(vii) Digital Image Classification
For estimation of area under agroforestry, choice of the satellite data pertaining to some specific
period in which agroforestry trees can be easily identified is very important. Field observation
with satellite imagery shows that in the study area, there may be two critical periods for choosing
the satellite data for delineation of agroforestry areas. During the kharif season, vegetation
growth is good and because of canopy density it gives a typical signature. This signature helps in
delineating agroforestry area from nearby cropland. It gives very typical blackish red signature
because of healthy vegetation. As per the crop calendar, the October month is ideal for acquiring
satellite image and during this period cloud free images are available for winter crop area
estimation. The September image for delineation of agroforesty is advisable as during this period
mature agroforestry field is in a unique pattern giving a unique spectral signature, which could be
easily delineated from rest of the neighboring land cover classes especially from other crops
grown during that period.
Classification Method
Maximum likelihood classifier was found to be accurate enough for classifying the study area for
extraction of agroforestry fields. In this supervised classification technique, training sites were
generated based on detailed ground survey.
The equation for the maximum likelihood/Bayesian classifier is as follows:
D ln ac 0.5 ln Covc 0.5 X M c Covc X M c
T 1
where
D = weighted distance (likelihood)
C = a particular class
IV.243
X = measurement vector of the candidate pixel
IV.244
and Kappa coefficient were obtained which indicates percent agreement between classified and
actual land uses/ land covers.
RAW DATA
LAYER STACK
EDGE MATCHING
MOSAIC IMAGES
SUBSET OF AOI
ACCURACY ASSESMENT
IV.245
5. Results and Discussions
The land use/land cover analysis of the district was done using both unsupervised and supervised
classification methods. Under unsupervised classification, nine land use/land cover classes viz.
cropland, agroforestry, scrubland and fallow land, reserve forest, plantation, water bodies, sand
and dry streams and built-ups were identified for the district. Area under these classes was
estimated and it was found that area under reserve forest and plantation are 12.27 and 15.10
percent respectively which is on higher side. Further, area under agriculture is only 18.42 percent
which is on lower side. The estimates for identified classes are shown in Table 2 which are not
found to be reliable.
Table 3: Land use/land cover in Ludhiana district using maximum likelihood classification method
Maximum Likelihood
Land use/ land cover classes Area (ha) Area (Sq.km.) Percentage
Cropland 3,08,439.85 3,084.39 83.70
Agroforestry 6,043.50 60.43 1.64
Current Fallow 1,142.36 11.42 0.31
Reserve Forest 1,449.01 14.49 0.39
IV.246
Plantation 325.86 03.25 0.08
Scrubland 133.90 01.33 0.03
Built-ups 29,717.30 297.17 8.06
River, Canal & Water bodies 1,949.61 19.49 0.53
Sand and dry streams 611.98 06.11 0.19
Unavailability of satellite data(restricted area) 18,497.00 184.97 5.07
Total 3,68,506.50 3,685.02
It was observed that sand & dry riverbed was intermingled with settlements. So in order to
reduce such error, both settlements and dry riverbed in the district were masked with the help of
toposheets and FCC. Similarly, agroforestry and forest were intermixed. For minimizing the
intermixing pixels, masking approach was applied to estimate area under agroforestry. The forest
area was masked out from the whole images of the district and hence area under forest was
obtained. Land use/land cover map for the district was generated.
Classification accuracy was tested in terms of Kappa co-efficient. Ground truth data on different
land uses including agroforestry was collected from the district and classification accuracy was
computed for the district using maximum likelihood method of supervised classification. Error
matrix generated using maximum likelihood method which is shown in Table 4 showed that
error of omission for agroforestry class is only 3.04 per cent, which means there is an accuracy
of about 96.96 per cent for this class. The overall accuracy of classification is 94.28 per cent
which is very significant and Kappa coefficient comes out to be 0.93, indicating 93 percent
agreement between classified and actual land uses/land covers.
Table 4: Error matrix for Ludhiana district using maximum likelihood classifier
AG CF AF SL WB DR RF PL BU Row Err.C
Total (%)
AG 36 01 01 00 01 01 00 00 02 42 14.29
CF 00 32 00 00 00 00 00 00 00 32 0.0
AF 00 00 27 00 00 00 00 00 00 27 0.0
SL 00 00 00 05 00 00 00 00 00 05 0.0
WB 00 00 00 00 31 01 00 00 00 32 3.12
DR 00 00 00 00 01 31 00 00 00 32 3.12
RF 00 00 00 01 00 00 33 00 00 34 2.94
PLN 00 00 00 01 00 00 00 26 00 27 3.71
T
IV.247
BU 02 00 00 00 00 00 00 00 31 33 6.07
Total 33 33 28 07 33 33 33 26 33 264
Err.O 6.07 3.04 3.58 28.58 6.07 6.07 0.0 0.0 6.07 5.72
(%)
IV.248
6. Sahoo, P.M., Rai, A., Singh, R., Handique, B.K. and Rao, C.S. (2005). Integrated
approach based on Remote Sensing and GIS for estimation of area under paddy crop in
North-Eastern Hilly region. Jour. Ind. Soc. Agril. Statist. 59 (2), 151-160.
7. Singh, R., Goyal, R.C., (2000). Use of remote sensing satellite data in crop surveys.
Project Report, IASRI, New Delhi.
IV.249
APPLICATION OF REMOTE SENSING AND GIS FOR
CROP NUTRIENT MANAGEMENT
K. N. Singh
I.A.S.R.I., New Delhi -110012
1. Introduction
With the global shift towards market economies, the need for timely and reliable agricultural
information has become more important than ever before in the decision making process. A
comprehensive knowledge of the basic soil resources is of fundamental importance for
efficient land use planning. The Indian subcontinent possesses widely divergent spectra of
changing physiography, climate and vegetation and their combined influence, accentuated by
the action of water and wind on the weathering of different types of parent materials, which
has obviously resulted in soils showing appreciable variations in morphological, physical,
chemical and biological characteristics. Thus the soil representing a continuum of diversified
genetic processes and being one of the biggest natural heritages of mankind deserves greater
consideration than merely as an inert medium of plant growth.
The acute problem of tremendously growing pressure of population in the country demands
maximum possible output of food and fiber from each unit of cultivated land area per unit
time. Soil fertility changes due to cropping, manure and fertilizer applications. Soil fertility
evaluation requires a unified approach. There is a series of steps involved. Results can only be
successful if the evaluation is considered as a process. Soil test results from one farm have to
be connected with the broader scope of the population of all farms in the given area. The ideal
would be to sample every farm. But we may not be able to sample each farm in the population,
because it is too costly and time consuming, especially with the multiple small farm holdings
in many developing countries. We need to generalize results over an area.
During the periods from 1975 to 1980, Dr. A. B. Ghosh and his associates from IARI have
prepared soil fertility maps for N, P and K using soil test data generated by soil testing
laboratories functioning throughout the country (Ghosh and Hasan, 1979). Till date there is no
major up-gradation in those maps, although the number of soil testing laboratories has
increased to around 520 and these laboratories have been generating soil test data of nearly 6.5
million samples per annum. The representative nature of soil samples and the quality of
analysis in soil testing laboratories is often questioned. So, there is a need to crosscheck these
maps through generating actual soil test data based on systematic soil sampling. Singh et al.
(2004) have used point estimates for districts to prepare soil fertility maps for N, P and K for
the states of Andra Pradesh and Maharastra and have interlinked these districts with
recommendations. Recently emerged technologies like Geographic Information systems (GIS),
RS and GPS have much to offer for preparing soil fertility maps. Patel et al. (2001) assessed
the land capability of Solani watershed of Haridwar and Saharanpur districts in Uttaranchal
and Uttar Pradesh, respectively to adopt suitable soil conservation measures and suggested
appropriate land use through remote sensing and geographical information system (GIS).
Bhuyan et al. (2002) has applied combined capabilities of remote sensing, geographic
information systems (GIS), and the agricultural non-point source pollution (AGNPS) model to
assess runoff and sediment yield from various sub-watersheds above Cheney Reservoir in
Kansas, USA. Ray and Dadhwal (2001) have used satellite-based remote sensing (RS) data and
geographic information system tools for estimating seasonal crop evapotranspiration in Mahi
Right Bank Canal (MRBC) command area of Gujarat, India.
Once the soil fertility maps are created, it is possible to transform the information from Soil
Test Crop Response models into spatial fertilizer recommendation maps. Such maps will
provide site-specific recommendation without testing the soil. The recommendations can be
obtained by an extension agent/farmer simply by locating his area on the map. Remote sensing
has evolved in to an important supplement to ground observations in the studies of terrestrial
vegetation. It is widely recognized that satellite remote sensing can provide an in expensive,
rapid and effective method of data collection and production of various kinds of thematic
maps. Remote sensing has been extensively used in the agriculture for vegetation studies.
Different types of Vegetation Indices have been used to study yield potential. Remote sensing
has been used to classify area under different crops and to forecast the yield levels of different
crops. It has been used to distinguish between different crops of similar signatures. It has also
been used to demarcate an area into different productivity zones. Preparation of soil fertility
maps districts wise using actual soil samples is very costly and time consuming. Remote
Sensing can provide speedy and inexpensive data for preparing these maps which can be used
for precise fertilizer recommendations based on available soil nutrients.
2. Material and Methods
Hoshangabad was selected for fertility mapping as it has highest consumption of fertilizer in
The list of villages including related information of the districts was obtained from the Census
of India (Bhopal). Hoshangabad district comprises of seven Tehsils (viz. Bankhedi, Bawai,
Hoshangabad, Itarsi, Pipariya, Seonimalwa and Sohagpur). As per our sampling plan
(Stratified multistage Random sampling) these Tehsils have been considered as strata. The
villages of each tehsil were further classified in to irrigated and un-irrigated villages with the
help of Deputy Director (Agriculture) Hoshangabad. Finally, 8% (Approx.) villages from each
group were selected using Random Number Table. From each selected village six farmers (two
of each categories viz. large (> 3ha), medium (1-3 ha) and small (< 1 ha)) were selected for
collecting further information. Proformas (for village as well as farmers) have been prepared
and multi-copied for final survey. Rural Agricultural Extension Officers (RAEO) filled the
village level proformas for each selected village. This proforma mainly contained the
information such as total cultivable land, crops and area, distribution of farmers according to
land, etc. This proforma was the base for final survey. Before actually reaching to a village
preliminary exercise such as the route map of the village was prepared. With the help of
Deputy Director (Agriculture) Hoshangabad, the concerned RAEO of the village was also
instructed to accompany us as a liaison between farmers and us. The survey work of the
Hoshangabad district was started in the month of May and finished by end of June. After
scanning of all the toposheets of Hoshangabad District we digitized the district boundary along
with tehsils boundaries. Mosaicing (merging of digitized boundary) was done and thorough
editing for error free data has been prepared. After assigning the point value (N, P, K, etc.) to
these sample points the kriging has been performed.
3. Model Selection
We are using Akaike’s (1973) information criterion (AIC) for selecting model, which can be
estimated as
Estimated (AIC) = n ln (R)+2p
Where,
n = The number of observations,
p = The number of estimated parameters,
R = The residual sum of squares of deviations from the fitted for model
The model having the smallest AIC is the best.
IV.250
4. Digitization of Toposheets and Georeferencing
Digitization of all the toposheets based on District boundary individually was done. After
digitization Mosaicing of (merging of digitized boundary) district boundary was completed.
The digital map with Tehsil boundaries of Hoshangabad district at the scale 1:50,000 along
with selected fields is given in Fig 1.
Figure 1. Digital map with Tehsil boundaries and Selected fields of Hoshangabad
district.
Figure 1 shows the sample fields (dots) in the Hoshangabad district and that some of the
selected fields are lying outside the district boundary. Similarly in case of tehsil some of the
selected fields are lying outside the tehsil boundary.
5. Result and Discussion
Generation of Response surface
IV.251
Estimated response surface (kriging) using Exponential method clearly shows that in
Hoshangabad district one can get soil O.C. in the range of 0.28% to 0.81% (Fig. 2). With the
help of this, output of all the ground points (pixel) has been assigned with unique estimated
value of O.C.%, which can be utilized for recommending its optimum use.
IV.252
case two or more adjacent grids are having same estimated values they are merged.
IV.253
Figure 7. Kriged raster (response surface) of electrical conductivity (dS/m)
Estimated response surface (kriging) using Linear method clearly showed that in
Hoshangabad district one can get electrical conductivity in the range of 0.086 to 0.345 micro
siemens (s) (Fig. 7). With the help of this, output of all the ground points (pixel) has been
assigned with unique estimated value of electrical conductivity. It is observed that the
electrical conductivity is high in the boundary areas of the river.
IV.254
Estimated response surface (kriging) using Linear method clearly showed that in
Hoshangabad District one can get pH in the range of 7.2 to 7.9 (Fig. 8). With the help of this,
output of all the ground points (pixel) has been assigned with unique estimated value of pH.
Fig 8 indicates that soils are mainly neutral to slightly alkaline in nature.
6. Validation of the prepared maps through post soil sample analysis
For the validation of prepared soil fertility maps which was prepared using the samples
collected in the summer of 2006, soil sampling was done in two consecutive years viz. 2007
and 2008. In 2007 two blocks in Hoshangabad district were selected randomly and seven soil
samples were randomly selected across the block from each block. In 2008 soil samples were
collected from all the blocks. After collecting soil samples these were analysed in the IISS
Central Lab for various soil properties. In this way we obtained a set of analysed data for
these two years. Similarly we obtained two set of estimated data through our prepared maps
(Raster images Figure 9-14). These samples were analyzed in the laboratory and the results
are given the Table 1. Paired t-test was used to test whether the pair of observations (x i, yi)
(where xi and yi are test and estimated nutrient values) corresponds to the same ith sample
unit. It is seen from Table2 that calculated Abs (t) is less that tabulated t (for P< 0.05) for all
the nutrients in 2007, which shows that in subsequent year there is no change in these
nutrients. If we see the results of year 2008, only for pH there is change. For other values
there is no any significant difference. Therefore, we may infer that that these parameters for
Hoshangabad district does not change significantly for at least two consecutive years except
pH.
IV.255
Table 1: Comparison of actual nutrient analyzed and its value through prepared maps (response surface)
IV.256
Table 2: Comparison of actual nutrient analyzed and its value through prepared maps
(response surface).
Year 2007
N (kg/ha) P (kg/ha) K(kg/ha) OC Ec (dS/m) pH
Mean (Analysed) 161.75 17.18 433.45 .60 .204 7.53
Mean (estimated) 154.68 17.72 429.72 .57 .197 7.60
df 13 13 13 13 13 13
t Stat 1.98 -1.38 .34 1.00 1.34 -1.21
t (Critical two-tail) for P<0.05 2.16 2.16 2.16 2.16 2.16 2.16
Year 2008
N (kg/ha) P (kg/ha) K(kg/ha) OC Ec (dS/m) pH
Mean (Analysed) 222.02 16.88 388.13 .62 .176 7.53
Mean (estimated) 218.38 16.64 390.95 .61 .173 7.62
df 85 85 85 85 85 85
t Stat -1.48 -1.46 .80 -1.23 -.98 4.12
t (Critical two-tail) for P<0.05 1.99 1.99 1.99 1.99 1.99 1.99
Satellite data (soft copy) for selected areas of Hoshangabad District was purchased from
NRSA, Hyderabad. Cloud free data for the months December 2005, January, February and
March 2006 was further processed for Hoshangabad District.
IV.257
Figure 11. Part of Hoshangabad District Figure 12. NDVI image of part of
obtained from the corrected scene Hoshangabad district on 29-1-2006
on 29-1-2006.
Figure 13. Part of Hoshangabad District Figure 14. NDVI image of part of
obtained from the corrected scene on 22-2- Hoshangabad district on 22-2-2006.
2006.
IV.258
Figure 15. Part of Hoshangabad District Figure 16. NDVI image of part of
obtained from the corrected scene on 18-3- Hoshangabad district on 18-3-2006.
2006.
It is clear from Figure 11 and 13 that the agricultural vegetation was at its full swing in the
months of Jan and Feb and Figure 15 shows that the crops of these areas have been
ripened/harvested from the months of March. Figure 10, 12, 14 and 16 represents NDVI
images for the months Dec, Jan, Feb and March respectively. Our aim is to find out
relationship between nutrients and NDVI values. To obtain the relationship between different
nutrients and NDVI values, following different forms of NDVI’s will be obtained:
1. Mean of mean NDVI’s values for different classes.
2. Mean of maximum NDVI’s values for different classes.
3. Maximum of mean NDVI’s values for different classes.
4. Maximum of maximum NDVI’s values for different classes.
Different types of linear and non-linear (such as linear, quadratic, cubic, logarithmic, power,
compound, inverse, exponential, logistic, growth etc.) relationship were evaluated.
To carry out statistical analysis the union was performed as below (Figure 17):
IV.259
Figure 17. District along with union of vector layers of NDVI and different nutrients (N,
P and K)
While performing unions we obtain frequency data for each nutrient polygons. This
frequency has also been used to improve the relationship. In this report only simple
correlation is being presented. The results indicate satisfactory relationship between nutrients
and NDVI values.
Table 3: Correlation Coefficient (r) between available Nitrogen and NDVI values for
different periods.
IV.260
Table 4: Correlation Coefficient (r) between available Phosphorus and NDVI values for
different periods.
Table 5: Correlation Coefficient (r) between available potassium and NDVI values for
different periods.
** P< 0.01
Table 3 clearly shows that there is a good agreement between available nitrogen and Max of
Max of NDVI values and it can be best estimated in the month of December and February.
Similarly Table 4 shows that available phosphorus can be estimated using Average of
Average of NDVI values in the month of February. Table 5 shows that Potassium can be
estimated using Average of Average of NDVI for the month of December. These relations
will further improve with higher order polynomials which will be presented in the final
report.
References
1. Akaike, H. (1973). Information theory as an extension of the maximum likelihood
principle. 267-281 in B. N. Petrov and F. Csaksi, editors. 2nd International
Symposium on Information Theory. Akademiai Kiado, Budapest, Hungary
2. Bhuyan, S.J., Marzen, L.J., Koelliker, J.K., Harrington, J.A., Barnes, P.L. (2002).
Assessment of runoff and sediment yield using remote sensing, GIS, and AGNPS.
Journal of Soil and Water Conservation Ankeny. 57 (6), 351-364.
3. Ghosh, A. B. and Hasan, R. (1979). Bulletin of Indian Society of Soil Science, 12, 1-
8.
4. Singh, K. N., Raju, N. S., Rao, A. S., Srivastava, S. and Maji, A. K. (2004). GIS based
system for prescribing optimum dose of nutrients for targeted yield through soil
fertility maps in Andhra Pradesh (AP). In the Proceedings of National workshop on
recent trends in earth resources mapping. 67-72.
IV.261
5. Singh, K. N., Raju, N. S., Rao, A. S., Srivastava, S. and Maji, A. K. (2004). GIS based
system for prescribing optimum dose of nutrients for targeted yield through soil
fertility maps in Maharastra. In the Proceedings of National Seminar on Information
and Communication Technology for Agriculture and Rural Development. 167-174.
6. Patel, N.R., Prasad, J., Kumar, S., Prasad, J. and Kumar, S. (2001). Land capability
assessment for land use planning using remote sensing and GIS. Agropedology. 11(1),
1-8.
7. Ray, S.S. and Dadhwal,V.K. (2001). Estimation of crop evapotranspiration of
irrigation command area using remote sensing and GIS. Agricultural Water
Management. 49(3), 239-249.
IV.262
APPLICATION OF REMOTE SENSING AND GIS IN
PRECISION FARMING
R. N. Sahoo
I.A.R.I., New Delhi -110012
1. Introduction
Historically, agronomic practices and management recommendations have been developed
for implementation on a field basis. This generally results in the uniform application of
tillage, fertilizer, sowing and pest control treatments at a field scale. Farm fields, however,
display considerable spatial variation in crop yield, at the 'within-field' scale. Such uniform
treatment of a field ignores the natural and induced continuous variation in soil properties,
and may result in zones being under- or over-treated, giving rise to economic and
environmental problems associated with this inefficient use of resource inputs. The more
substantial of these problems being: economically significant yield losses, excessive chemical
costs, gaseous or percolatory release of chemical components, unacceptable long-term
retention of chemical components and a less than optimal crop growing environment.
Precision Farming (PF), in the form of Site-Specific Management, offers a remedy to many
of these concerns. The philosophy involves matching resource application and agronomic
practices with soil properties and crop requirements as they vary across a site. PF has three
requirements such as (i) ability to identify each field location, (ii) ability to capture, interpret
and analyze agronomic data at an appropriate scale and frequency, and (iii) ability to adjust
input use and farming practices to maximize benefits from each field location. Collectively,
these actions are referred to as the "differential" treatment of field variation as opposed to the
"uniform" treatment underlying traditional management systems. The result is an
improvement in the efficiency and environmental impact of crop production systems. It is
sometimes called ‘Prescription farming’, site-specific farming’ or ‘variable rate technology
farming’.
2. Developments which Prompted Precision Farming
Many technological developments occurred in 20th century led to the development of the
concept of precision farming. The success of the precision farming system relies on the
integration of these technologies into a single system that can be operated at farm level with
sustainable effort. These technological developments are as follows.
2.1 Computer and Internet
The computers and internet are the most important component in enabling the precision
farming possible as they are main source of information processing and gathering. The high-
speed computer has made faster processing the data gathered during precise management of
the land parcel. Internet, which is a network of computers, is the most recent development
among all these technologies. The Internet has bridged the gap between the information
provider and the user. In agriculture, like any other form of business, internet has the
capability to supply timely data about changing conditions.
2.2 Global Positioning System
The development of the publicly available global positioning system (GPS) has opened new
doors in opportunities for spatial data. This is a passive positioning system from a
constellation of 24 orbiting radio-navigation satellites. They provide continuous position data
in two or three dimensional in real-world coordinates.
More recently farmers have gained access to site specific technology though GPS. GPS
makes use of a series of satellites that identify the location of farm equipment within a meter
of an actual site in the field. The most common use of GPS in agriculture is for yield mapping
and variable rate fertilizer/pesticide applicator. GPS are important to find out the exact
location in the field to assess the spatial variability and site-specific application of the inputs.
GPS operating in differential mode are capable of providing location accuracy of 1 m.
The availability of GPS approaches to farming will allow all field-based variables to be tied
together. This tool has proven to be the unifying connection among field variables such as
weeds, crop yield, soil moisture, and remote sensing data.
2.3 Remote Sensing Technique
Remote sensing (RS) holds great promise for precision agriculture because of its potential for
monitoring spatial variability over time at high resolution (Moran et.al., 1997). Various
workers (Hanson et al., 1995) have shown the advantages of using remote sensing technology
to obtain spatially and temporally variable information for precision farming. Remote sensing
imagery for precision farming can be obtained either through satellite-based sensors or CIR
video digital cameras on board small aircraft. Moran et al. (1997) in their review paper
summarized the applications of remote sensing for precision farming. They have found RS
can be used as source of different types of information for precision farming. However, using
RS data for mapping has many inherent limitations, which includes, requirements for
instrument calibration, atmospheric correction, normalization of off-nadir effects on optical
data, cloud screening for data especially during monsoon period, processing images from
airborne video and digital cameras (Moran et al., 1997). Keeping in view the agricultural
scenario in developing countries, the requirement for a marketable RS technology for
precision agriculture is the delivery of information with the following characteristics like low
turn around time (24-48 hrs), low data cost (~ 100 Rs./acre/season), high spatial resolution (at
least 2m multi-spectral), high spectral resolution (<25 nm), high temporal resolution (at least
5-6 data per season) and delivery of analytical products in simpler format.
2.4 Geographic Information System (GIS)
GIS is a computerized data storage and retrieval system, which can be used to manage and
analyze spatial data relating crop productivity and agronomic factors. It can integrate all types
of information and interface with other decision support tools. GIS can display analyzed
information in maps that allow (a) better understanding of interactions among yield, fertility,
pests, weeds and other factors, and (b) decision-making based on such spatial relationships.
Many types of GIS software with varying functionality and price are now available. A
comprehensive farm GIS contains base maps such as topography, soil type, N, P, K and other
nutrient levels, soil moisture, pH, etc. Data on crop rotations, tillage, nutrient and pesticide
applications, yields, etc. can also be stored. GIS is useful to create fertility, weed and pest
intensity maps, which can then be used for making maps that show recommended application
rates of nutrients or pesticides.
2.5 Spatial Decision Support Systems (SDSS)
Spatial decision support systems (SDSS) are designed to help growers to solve complex
spatial problems and to make decision concerning to irrigation scheduling, fertilization, use
of crop growth regulators and other chemicals. Spatial decision support systems have evolved
in parallel with decision support systems (DSS). In addition, in order to effectively support
decision- making for complex spatial problems, a SDSS will need to:
IV.264
i. provide for spatial data input
ii. allow storage of complex structures common in spatial data
iii. include analytical techniques that are unique to spatial analysis
iv. provide output in the form of maps and other spatial forms
GISs stand alone fall short of the goals of SDSS for a number of reasons:
i. analytical modeling capabilities often are not part of a GIS
ii. many GIS databases have been designed solely for cartographic display of
results- SDSS goals require flexibility in the way information is
communicated to the user
iii. the set of variables or layers in the database may be insufficient for complex
modeling
iv. data may be at insufficient scale or resolution
v. GIS designs are not flexible enough to accommodate variations in either the
context or the process of spatial decision-making
SDSS provide a framework for integrating: 1) analytical modeling capabilities 2) database
management systems 3) graphical display capabilities 4) tabular reporting capabilities 5) the
decision-maker's expert knowledge. GISs normally provide 2), 3) and 4).The addition of 1)
and 5) create a SDSS.
2.6 Yield Mapping
Yield mapping and soil sampling tend to be the first stage in implementing PF. Yield maps
are produced by processing data from an adapted combine that has a vehicle positioning
system integrated with a yield recording system. Massey Ferguson were the first company to
produce a commercial yield mapping combine. This combine has a Differential Global
Positioning System (DGPS) fitted to it that can be identified by the GPS receiver on the roof
of the cab and the differential aerial above the engine. The output from the combine is a data
file that recorded every 1.2 seconds the position of the combine in longitude and latitude,
with the yield at that point. This data set can then be processed by various geo-statistical
techniques into a yield map.
2.7 Crop Simulation Models (CSMs)
However, although significant technical advances have been made in measuring and
displaying variation in crop yields across a field, it is not always clear how to determine the
best management practice for each part of a field in order to achieve these goals. For
example, does a farmer apply more fertilizer to the lower yielding areas to try and raise yields
to the average or are these low-yielding areas at their potential yield already, and would he
therefore be advised to apply more to the high yielding areas, believing they are able to make
better use of it. Answers to these questions are not clear cut, but can be found using
optimization techniques based on knowledge of the marginal yield response to a unit of the
input in question for all the different regions of a field. The marginal yield response can be
obtained from the curve describing yield response to the level of particular input for each
homogenous region are unlikely to be known. While mini-experiments on each homogenous
region with a range of application levels of the input will provide this information, this
approach in time-consuming, labour-intensive, and results are likely to be specific to that
field only. An alternative, and perhaps complementary, approach is to use crop simulation
models to predict the likely yield response to different levels of a particular input. Such
models offer a cost effective way in which agronomic knowledge accumulated from
IV.265
numerous previous experiments, usually with treatments of uniformly applied inputs on small
and relatively homogenous plots, can be extended to larger spatially-variable fields.
Crop simulation models are needed to help consultants, researchers, and other farm advisors
determine the pattern of field management that optimizes production or profit. However, the
effective use of these tools requires their evaluation in fields to be optimized, their integration
with other information tools such as GIS, geostatistics, remote sensing, and optimization
analysis.
Crop simulation models like CERES (maize, wheat, rice, sorghum, barley, and millet)
CROPGRO (soybean, peanut, dry bean, and tomato), SUBSTOR (potato), CROPSIM
(cassava), and CANEGRO (Sugarcane) models has been developed by researchers from
several countries. These models respond to weather, soil water holding and root growth
characteristics, cultivar, water management, nitrogen management, and row spacing/plant
population. Also decision support system like, DSSAT incorporates crop/soil/weather
models, data input and management software, and analysis programs for optimizing
production or profit for homogenous fields. DSSAT also includes links to GIS and remote
sensing information, which allows mapping of spatially variable inputs across a field and
mapping of predicted outputs from the models, such as yield, nitrogen leaching, water use,
etc. The site specific yield potentials can be estimated determining spatial pattern crop and
land information and using it in above simulation models.
2.8 Variable Rate Technology
The variable rate technology (VRT) is the most advanced component of PF technologies,
provides "on-the-fly" delivery of field inputs. A GPS receiver is mounted on a truck so that a
field location can easily be recognized. An in-vehicle computer, which contains the input
recommendation maps, controls the distribution valves to provide a suitable input mix by
comparing to the positional information received from the GPS receiver. Current commercial
VRT systems are either map-based or sensor-based (NRC, 1997). The map-based VRT
systems require a GPS/DGPS geo-referenced location and a command unit that stores a plan
of the desired application rates for each field location. The sensor-based VRT systems do not
require a geo-referenced location but include a dynamic control unit, which specifies
application through real time analysis of soil and/or crop sensor measurements for each field
location. New VRT systems like the manure applicator being developed at Purdue University
may soon enable precise application of manure in cropping systems. There are two methods
of VRT. The first method, Map-based, includes the following steps: grid sampling a field,
performing laboratory analyses of the soil samples, generating a site-specific map of the
properties and finally using this map to control a variable-rate applicator. During the
sampling and application steps, a positioning system, usually DGPS (Differential Global
Positioning System) is used to identify the current location in the field. The second method,
Sensor-based, utilizes real-time sensors and feedback control to measure the desired
properties on-the-go, usually soil properties or crop characteristics, and immediately use this
signal to control the variable-rate applicator.
3. Precision Farming Components and Framework
Precision Farming, basically, is characterized by reduced cost of cultivation (through
optimization of inputs), improved control and increased resource use efficiency, through
appropriate applications of Management Information System (MIS). While the reduced cost
of cultivation is achieved through optimization of agricultural inputs taking into account
economic push and environmental pull related factors, the control mechanisms are introduced
by the help of VRT systems, model outputs and conjunctive use of remote sensing, GIS and
IV.266
GPS. The MIS comprises Decision Support Systems (DSS), collateral inputs and associated
GIS databases on crops, soils & weather. Dynamic remote sensing inputs on in-season crop
conditions, crop simulation model outputs on the potential production under the different
constraining scenario, and the networks of labs and farms, form the essential ingredients of
MIS. Increased efficiency does not employ only efficient resource use but also reflects in
terms of less waste generation, improved gross margin and reduced environmental impact.
Precision Farming thus calls for the use of appropriate tools and techniques, within a set of
the framework as mentioned, to address the micro-level variations between crop requirements
and applications of agricultural inputs. Inevitably, it integrates a significant amount of data
from different sources; information and knowledge about the crops, soils, ecology and
economy but higher levels of control require a more sophisticated systems approach. It is not
simply the ability to apply treatments that are varied at the local level but the ability to
precisely monitor and assess the agricultural systems at a local and farm level. This is
essentially to have sufficient understanding of the processes involved to be able to apply the
inputs in such a way as to be able to achieve a particular goal not necessarily maximum yield
but to maximize financial advantage while operating within environmental constraints.
4. Precision Farming in India
PF technologies may be relatively new to India, but the concept of precision management is
not. Indian farmers have long known that soil conditions, fertility, moisture, etc. vary widely
across a single field and that various parts within fields responded to different types of inputs,
and cultural practices. The small size of their farms often permitted such an effective
monitoring of spatial and temporal yield variation and variable application of inputs mainly
by manual means.
4.1 Limitations
Although precision farming is a proven technology in many advanced countries of the world
but its scope in India (also in other developing countries) are limited. Different scientists have
reported certain constraints, which limited the scope for site-specific farming in India, are as
given follows:
i. Small farms size, heterogeneity of cropping systems, and land tenure/ownership
restrictions High cost of obtaining site-specific data
ii. Complexity of tools and techniques requiring new skills
iii. Culture, attitude and perceptions of farmers including resistance to adoption of new
techniques and lack of awareness of agro-environmental problems
iv. Infrastructure and institutional constraints including market imperfections
v. PF as new story to Indian farmers needs demonstrated impacts on yields
vi. Lack of local technical expertise
vii. High initial investment
viii. Uncertainty on returns from investments to be made on new equipment and
information management systems, and
ix. Knowledge and technological gaps including
Inadequate understanding of agronomic factors and their interaction,
Lack of understanding of the geostatistics necessary for displaying spatial
variability of crops and soils using current mapping software, and
Limited ability to integrate information from diverse sources with varying
resolutions and intensities.
IV.267
4.2 Opportunities
Though farm size is the major limitation in adoption of PF technologies in India, contiguous
field with same crop and mostly under similar management practice in states like Haryana,
Punjab and Rajasthan can be considered as potential site for precision farming. Punjab and
Haryana states in India, where farm mechanization is more common than in other states, are
likely to be the first to adopt precision farming on a large scale. Rice, wheat, sugar beet,
onion, potato, and cotton among the field crops, and apple, grape, tea, coffee and oil palm
among horticultural crops are perhaps the most relevant for precision farming. Some have a
very high value per acre, making excellent cases for site specific management. For all these
crops, yield mapping is the first step to determine the precise locations of the highest and
lowest yield areas of the field, and to analyze the factors causing yield variation.
In India, a few researchers in the private sector initiated studies on precision agriculture in
high value crops like cotton, coffee and tea. In cotton, remote sensing coupled with GIS can
assist in improved precision of insect pest management and harvesting. Testing of precision
farming technologies for sustainable rice and wheat cropping system at the research farm
level is in progress at Indian Agricultural Research Institute, New Delhi. Space Application
Centre (SAC, ISRO), Ahmedabad in collaboration with Central Potato Research Institute,
Simla has started experiment in Central Potato Research Station Farm at Jalandhar, Punjab to
study the role of remote sensing in mapping the variability. The National Bank for
Agriculture and Rural Development (NABARD) with collaborative partnership with MS
Swaminathan Foundation, Chennai and Arava R&D Centre in Israel has established Resource
Centre for Precision Farming for developing and spreading production technologies based on
integrated natural resources management. Project Directorate of Cropping System Research,
Modipuam, in collaboration with SAC and National Remote Sensing Centre (NRSC) in
collaboration with ICRISAT, CRIDA and ANGRAU are involved in precision farming
research for capturing variability and variable rate input application.. There is also great
opportunity of PF for grape and tea to start as pilot project in Nashik district of Maharastra
and Assam respectively, where these crops grown in concentrated area.
Nutrient and water stress management is another area where precision farming can help
Indian farmers. Detecting nutrient stresses using remote sensing and combining data in a GIS
can help in site-specific applications of fertilizers and soil amendments such as lime, manure,
compost, gypsum, and sulfur, which in turn would increase fertilizer use efficiency and
reduce nutrient losses (Sawyer, 1994). In semi-arid and arid tropics, precision technologies
can help growers in scheduling irrigation more profitably by varying the timing, amounts and
placement of water. For example, drip irrigation, coupled with information from remotely
sensed stress conditions (e.g., canopy-air temperature difference), can increase the effective
use of applied water thereby reducing runoff and deep percolation (Das and Singh, 1989).
Pests and diseases cause huge losses to crops. If remote sensing can help in detecting small
problem areas caused by pathogens, timing of applications of fungicides can be optimized.
Recent studies in Japan show that pre-visual crop stress or incipient crop damage can be
detected using radio-controlled aircraft and near-infrared narrow-band sensors. Likewise,
airborne video data and GIS have been shown to effectively detect and map black fly
infestations in citrus orchards, making it possible to achieve precision in pest control (Everitt
et al., 1994). Perennial weeds, which are usually position-specific (Wilson and Scott, 1982)
and grow in concentrated areas, are also a major problem in precision agriculture. Remote
sensing combined with GIS and GPS can help in site-specific weed management.
Although thorough cost-benefit analysis has not been done yet, the possible use of precision
technologies in managing the environmental side effects of farming and reducing pollution is
IV.268
appealing. The technologies need not be limited to input application. They can be used in (a)
implementing spatially-varied farm operations such as tillage, seeding, harvesting, etc., (b)
on-farm testing of agronomic practices to evaluate alternative management practices, (c)
plant breeding programs to test the performance of improved varieties, and in (d) re-
evaluations of trial procedures.
4.3 Strategies
Precision farming needs to go from a technology-push to application-driven approach. As no
single agency can take on the entire PF process, it is essential that various agencies join
together and give a participatory approach for effective implementation of PF technologies.
Small farm size will not be a major constraint, if the technologies are available through
consulting, custom and rental services. For instance, by renting the equipment, manufacturers
can enable early adapters to avoid the risks associated with purchasing costly machinery.
The role of agricultural cooperatives is important in dissemination of precision farming
technologies to small farmers. If precision farming is considered a series of discrete services:
map generation, targeted scouting, etc., it is possible to fit these services within the structure
of a progressive agricultural cooperative. Since PF technology requires many costly
implements farmers’ cooperative no the single farmer can afford to procure them. Again,
cooperative farming will solve the limitation of small field size to take PF.
Changes in agricultural policies are also necessary to promote the adoption of precision
farming. There are basically two policy approaches: regulatory policies, and market-based
policies. The former refer to environmental regulations on the use of farm inputs, and the
latter refer to taxes and financial incentives aimed at encouraging growers to efficiently
utilize farm inputs. In India, the lack of penalties for pollutant generation has partly
contributed to an excessive use of inputs. Subsidies on inputs and outputs, and mechanisms
that prevent the price system from rationing limited resources are also common. The latter
include state-guaranteed crop prices, tariffs, import quotas, export subsidies, etc.
5. Conclusion
Precision farming is essential for serving dual purpose of enhancing productivity and
reducing ecological degradation. The success stories pertaining to Precision Agriculture have
mainly drawn from the developed countries; wherein Agriculture is characterized by highly
mechanized and automated systems, and is driven by market forces and has been
professionally managed enterprise. Taking into account the predominance of fragmented land
holdings, heterogeneity of crops and livestock and concept of farm families in the rural
conditions, the model of Precision Agriculture representing the typical Indian Agricultural
scenario is yet to evolve. Although it is recognized that agriculture is a major polluter of the
environment, farmers will not adopt precision farming unless it brings in more or at least
similar profit as compared to traditional practice. While the ecological integrity of farming
systems is an imperative need, it is equally important to extend the access of information and
market to the small and marginalized farmers. The Precision Agriculture model for India
while addressing these issues would provide an innovative route for sustainable agriculture in
globalised and liberalized economy.
References
1. Blackmore, S. (1994). Precision farming: An introduction. Outlook on Agril., 23(4), 275-
280.
2. Byerlee, D. (1994). Modern varieties, productivity, and sustainability: Recent experience
and emerging challenges. CIMMYT, Mexico D.F., Mexico.
IV.269
3. Das, D.K. and Singh, G. (1989). Estimation of evapotransipiration and scheduling
irrigation using remote sensing techniques. Proc. Summer Inst. on agricultural remote
sensing in monitoring crop growth and productivity , IARI, New Delhi, 113-117.
4. Everitt, J.H., Escobar, D.E., Summy, K.R. and Davis, M.R. (1994). Using airborne video,
global positioning systems and geographic information system technologies for detecting
and mapping citrus blackfly infestations. Southwestern Entomol. 19(2), 129-138.
5. Hanson, L.D., Robert, P.C. and Bauer, M. (1995). Mapping wild oats infestation using
digital imagery for site specific management. In Proc. Site-Specific Mgt. for Agric. Syst.
27-230, March, 1994, Minneapolis, MN, ASA-CSA-SSSA, Madison, WI, 495-503.
6. Hatanaka, T., Nishimuna, A., Nira, R. and Fukuhara, M. (1995). Estimation of available
moisture holding capacity of upland soils using Landsat TM Data. Soil Sci. Plant Nutr.
41(3), 577-586.
7. Iida, M., Umeda, M., Kaho, T., Lee, C.K. and Suguri, M. (1998). Measurement of grain
yield in Japanese paddy field. p. 54. In Abstracts of the Fourth Int. Conf. Precision
Agriculture, July 19-22, 1998. St. Paul, MN.
8. Larscheid, G. and Blackmore, B.S. (1996). Interactions between farm managers and
information systems with respect to yield mapping. 1153-1163. In P.C. Robert et al.,
(ed.) Precision Agriculture. Proc. Third Int. Conf. Minneapolis, MN. June 23-26, 1996.
ASA, CSSA, SSSA, Madison, WI.
9. Maji, A.K., Sidhu, G.S., Kandpal, B.K., Pande, S., Velayudham, M. and Ahmed, M.I.
(1997). Evaluation and Management of Rice-wheat cropping systems in Indo-Gangetic
Plains using GIS. In Use of GIS in analysis of cropping systems, Training Workshop,
Patancheru. 20-29 Aug. 1997. ICRISAT, Patancheru, India.
10. Ministry of Agriculture. (2000). Agricultural statistics at a glance. Directorate of
Economics and Statistics. Dept. Agric. Coop., Gov. India. New Delhi, India.
11. Moran, M.S., Inoue, Y. and Barnes, E.M. (1997). Opportunities and limitations for
image based remote sensing in precision crop management. Remote Sensing
Environment, 61, 319-346.
12. Nagarajan, S., Dadhwal, V.K. and Sahoo, R.N. (2004). Role of earth observation
technology and applications in India’s food security Concerns, Proc. on India-United
States Conference on Space Science, Application and Commerce – Strengthening and
Expanding Co-operation, held at Bangalore, June 21-25, 2004, 25.
13. NRC, (1997). Precision Agriculture in the 21st Century Geospatial and information
techniques in crop management. National Academy Press, Washington DC. 149.
14. Pearce, D., Barbier, E. and Markanda, A. (1988). Sustainable development and cost-
benefit analysis. Paper 88-01. London Environmental Economics Centre, UK.
15. Pierce, F.J. and Nowak, P. (1999) Aspects of precision agriculture. Adv. In Agronomy,
67, 1-85.
16. Ray, S.S., Panigrahy, S and Parihar, J.S. (2001). Remote Sensing for Precision farming
with special reference to Indian situation, In scientific note.
SAC/RESA/ARG/AMD/SN/01/ 2001, Space Application Centre, Ahmedabad.
17. Sahoo, R.N., Tomar, R.K., Pandey, S., Sahoo, P.M., Chakaborty, D and Kalra, N. (2007)
Precision Farming: Concept and Application in Indian Context, Indian J Crop Sci 2(1):
25-27.
IV.270
18. Sahoo, R.N. (2003). Precision Farming: Concept and Approaches, In Proceeding of 1n
DOS funded 12th Winter School on Remote Sensing Application in Agriculture with
special emphasis on Precision Farming 2003:1-10.
19. Sahoo, R.N., Tomar, R.K. and Arora, R.P. (2002). Precision farming: A prospective in
21st century. Proc. 2nd Int. Agronomy Congress on Balancing food and environmental
security – A continuing challenge, IARI, New Delhi, Nov. 26-30, 2002, 1194-95.
20. Sahoo, R.N. and Arora, R.P. (2001), Precision farming in Indian agriculture. In Proc.
Natl. Symp. on ‘Recent Advances in Remote Sensing and GIS Technologies for Natural
Resources Management’ IIT, Mumbai, December 5-7, 2001: 1-10.
21. Sadler, E.J., Busscher, W.J., Bauer, P.J. and Karlen, D.J. (1998). Spatial scale
requirements for precision farming: A case study in the Southeastern USA. Agron. J.,
90,191-197.
22. Sawyer, J.E. (1994). Concepts of variable rate technology with considerations for
fertilizer application. J. Prod. Agric. 7(2), 195-201.
23. Shropshire G., Peterson, C. and Fisher, K. (1993). Filed experience with differential
GPS, American Society of Agricultural Engineers. Paper: 93 1073.
24. Srinivasan, A. (1998). Precision farming in Asia: Progress and prospects In: proceedings
of the 4th International Conference on precision agriculture, 19-22 July, 1998 St. Paul,
USA.
25. Swaminathan, M.S. (2000). Natural resource management – for an ever green revolution.
In the Hindu- Survey of Indian Agriculture, pp. 9-16.
26. Tabor, J. and Hutchinson, C. (1994). Using indigenous knowledge, remote sensing and
GIS for sustainable development. Indigenous knowledge and Development Monitor 2(1),
2-6.
27. Tanaka, A. (1997). Systemic production of horticultural plants by microcomputers. III.
Operation with LAN system for a production organization. (In Japanese). Agric. Info.
Res. 6(1), 9-25.
28. Wang, M.H. (2001). Possible adoption of precision agriculture for developing countries
at the threshold of the new millennium. Computers and Electronics in agriculture. 30, 45-
50.
29. Wilson, B.J., and Scott, J.L. (1982). Population trends of Avena fatua and Alepecuras
myosurooides on commercial arable and dairy farms. 619-628. In Weeds, Proc. British
Crop Prot. Conf. SCI, London, UK.
IV.271
APPLICATION OF GIS IN SPATIAL SAMPLING FOR
AGRICULTURAL SURVEYS
Prachi Misra Sahoo
I.A.S.R.I., New Delhi -110012
1. Introduction
Agricultural statistics plays an important role in national planning and judicious allocation of
limited resources to the different sectors of economy in the country. The domain of
agricultural statistics was initially based on two parameters the area and production
estimation of principal crops. The former is obtained through complete enumeration whereas
the latter through sample surveys. The recent technological developments in the computer
and space technology have shifted the emphasis of survey research work towards newer
emerging areas of Remote Sensing technology and Geographical Information System (GIS).
These advances in technology have given a new direction to the survey research including
agricultural surveys. Remote Sensing in the form of satellite data, emerged as a powerful tool
for agricultural surveys particularly in the field of area sampling. Advances in computer
geographic techniques have the potential to substitute the traditional sampling designs to the
GIS assisted sampling strategy. Modern applications of GIS technology in surveys
demonstrated its potential not only in terms of cost efficiency, but also for maintaining same
level of accuracy and timeliness. However, the variables used in agricultural surveys are
often geographical in nature, which implies the dependency in the data. Consequently, the
data exhibits spatial autocorrelation. Thus, an improvement over the existing spatial sampling
techniques for agricultural surveys, incorporating spatial autocorrelation and making use of
GIS was required.
Sahoo et al. (2006) proposed GIS assisted spatial sampling techniques. These techniques are
developed taking into account the dependency present in the geographical data. The extent of
dependency is related to the extent of spatial correlation. The spatial autocorrelation is then
used to assign weights, which in turn are used for calculating the probability of selection for
the sampling units. The weights are assigned in such a way that, the units which are far apart
from the earlier selected units gets the higher probability of selection. Another problem of
geographic units is irregularity in their shapes and sizes. This has been taken care of by
modifying the proposed technique for irregular data. Further, to improve the performance of
proposed GIS assisted spatial sampling techniques, remote sensing satellite data is introduced
in the form of auxiliary character. In this way, advantage of remote sensing technique has
been exploited and an integrated GIS and remote sensing technique is proposed. Further, to
judge the performance of proposed spatial sampling techniques a simulation study is
conducted for the crop area estimation in the Rohtak district of Haryana. The proposed spatial
sampling techniques are compared with the existing sampling designs recommended for
spatial data like Simple random sampling with replacement, Simple random sampling without
replacement, Stratified sampling and Dependent Areal Unit Sequential Technique (DUST)
proposed by Arbia (1993). The comparison is also made among the proposed techniques for
regular area units and irregular area units. Lastly, the spatial sampling technique is applied by
integrating both remote sensing and GIS to estimate area under wheat for the Rohtak district
and the estimates obtained are compared with the true value. A brief introduction to some
basic concepts in spatial statistics, GIS and Remote Sensing are mentioned in the subsequent
sections:
2. Some Concepts of Spatial Statistics
2.1 Spatial Sampling
Kotz and Johnson (1986) defined spatial sampling as that area of survey sampling which is
concerned with sampling in two dimensions like the sampling of fields, groups of contiguous
quadrats or other planar surface. One approach to spatial sampling is through a population of
MN units, usually points or quadrats, arranged in M rows and N columns. The sampling
designs to choose mn units fall into three distinct types: designs in which the sample units are
aligned in both the rows and column directions; designs in which the sample units are aligned
in one direction only, say the rows, and unaligned in column directions; designs in which the
sample units are unaligned in both the directions. For designs that have the sampling units
aligned in both the directions, the number of sample elements in any row of the population
will be 0 or n and the number of sampled elements in any column of the population will be 0
or m. For designs that have sampled units aligned in the rows and unaligned in column, the
number of sample elements in any row of the population will be 0 or n and the number of
sampled elements in any column will be at most m. Designs that have sample units unaligned
in both the directions are characterized by having at most n sample elements in any row and
at most m elements in any column of the population with the exception of simple random
sampling without replacement of mn units from the MN in the population. Three traditional
sampling designs have generally been applied for selection of the sampling units in different
ways such as (a) simple random sample of row/columns, (b) a stratified sample of rows and
for each selected row independent stratified sample of columns and (c) a systematic sample
unaligned in both the directions. The most commonly used designs are Simple random
sampling, unaligned stratified sampling, aligned systematic sampling and stratified
systematic unaligned sampling. Hedayat et al. (1988a) proposed balanced sampling plan
which excludes the selection of contiguous units to the earlier selected units. Hedayat et al.
(1988b) revealed that if there exist some ordering of units under which contiguous units are
anticipated to provide similar data, in that situation, more information on population could be
obtained if the sample avoids pair of contiguous units. The required ordering of units in these
situations can be induced by a natural entity such as time or locations. It may be an ordering
in one or more dimension. They introduced a class of sampling designs. Arbia (1990, 1991),
following the lines of Hedayat et al. (1988 a and b), proposed a draw by draw sampling
procedure for more efficient area sampling design viz. Dependent Areal Unit Sequential
Technique (DUST). It is a GIS based sequential technique characterized by variable
inclusion probabilities at each step.
2.2 Spatial Autocorrelation
Given a group of mutually exclusive units or individuals in a two dimensional plane, if the
presence, absence or degree of a certain characteristic affects the presence, absence or degree
of the same characteristic in neighbouring units, then the phenomenon is said to exhibit
spatial autocorrelation (Cliff and Ord, 1973). Spatial autocorrelation tests whether or not the
observed value of a variable at one locality is independent of values of that variable at
neighbouring localities. A positive spatial autocorrelation refers to a map pattern where
geographic features of similar value tend to cluster on a map, whereas a negative spatial
autocorrelation indicates a map pattern in which geographic units of similar values scatter
throughout the map.
2.3 Neighbourhood Analysis
The definition of neighbourhood is of prime importance, as alternative definitions can lead to
very different neighborhoods and hence, to different statistics (Unwin and Unwin, 1998).
IV.273
Several ways of defining a neighbourhood are possible. For point data, distance is the basis
for the definition of a neighbourhood. In case of areal data, there are many options among
which the most popular is contiguity/adjacency neighbours. Adjacency heavily depends on
size and shape of areas under investigation. Another way is defining distance-based
neighbourhood. The definitions for contiguity and distance-based neighbours are given
below.
Contiguity based Neighbors
I. Rook’s definition of contiguity:
If the data is organized as contiguous quadrats the Rook’s definition of contiguity
implies that quadrats are considered to be neighbours only if they are touching edges.
The quadrats are considered to be neighbours only if they are touching corners.
In many polygonal data sets, and for all point data, a natural definition of neighbours can be
based on metric distance. In the case of point data this is easily evaluated, but for the area
objects a point for each areal unit has to be identified. This can be the centroid of each areal
unit. Once the centroid is identified, an optimum distance is specified according to certain
criteria. All the areal units whose centroid falls within that distance are considered as the
neighbours of the unit from whose centroid distance is measured. Fig.2 shows that all the
areal units whose centroids fall within the smaller circle will be the neighbors of the central
areal unit of the circle.
IV.274
Figure 2. Distance based neighbours
Ln Ln Ln Ln Ln Ln Ln Ln Ln
: : : : :
Ln Ln
Ln ..
L2 L2 L2 L2 L2 ..
Ln C: central pixel
.. ..
Ln L2 L1 L1 L1 L2 Ln L1: 1st lag neighbours
.. ..
Ln L2 L1 C L1 L2 Ln L2: 2nd lag neighbours
.. ..
Ln L2 L1 L1 L1 L2 Ln Ln: nth lag neighbours
.. ..
Ln L2 L2 L2 L2 L2 Ln
.. : : : : :
Ln Ln
Ln Ln Ln Ln Ln Ln Ln Ln Ln
IV.275
3.1 Estimation of Spatial Correlation
Spatial dependence is measured by estimating the spatial correlation of the auxiliary variable
X at various spatial lags. Estimation of the spatial correlation requires the identification of
various order neighbours for each areal unit. The neighbours are identified using Queen’s
method, which has been described earlier in the first chapter. Once the neighbours are
identified the spatial correlation is calculated for X using the following formula given by
Moran (1950) :
w ij X i X X j X
N N
N i 1 j1
β … (3.1)
C
X i X
N 2
i 1
where,
N N
C w ij i j
i 1 j1
= 0, otherwise
3.2 Stationarity Test of Spatial Correlation
After the calculation of spatial correlation, zones are to be formed such that the spatial
correlation is homogeneous within these zoneswhich implies that the spatial correlation of all
the units falling within one zone is stationary. For this purpose a Monte Carlo significance
test given by Brunsdon et al. (1998) is applied.
3.3 Method of Sample Selection
Let there be N areal units out of which a sample of size n is to be selected. Further, let X be
the size measure, the value of which is known for each unit. The first lag spatial correlation is
denoted by . Let i denote the probability of selection of ith unit in the sample. The first unit
is selected with probability proportional to size sampling. The subsequent units are selected
drawn sequentially by assigning weight varying at each step. The weight for the selection of
units is taken to be Ui n (1 β
d1n
)(1 βd 2n )........... 1 β
d (n -1)n
X
in where d1n = 1, if 1st
and nnd units selected are 1st lag neighbour = 2, if 1st and nnd units selected are 2nd lag
neighbour and so on.
3.4 Estimation Procedure
Two estimators are proposed by Sahoo et al. (2006) which are given in the subsequent
sections.
IV.276
3.4.1 Contiguous Unit Based Spatial Sampling (CUBSS)
Let Y be the character under study taking values Yi on i = (1, 2,…,N). Let y1 , y 2 ,…, y n be
the values of the unit drawn at the first, second… nth draw respectively. Let s 1* , s*2 ,…, s *n
denote the sample set which contains the units selected from the N units after first, second,
…,n draws respectively, such that, s 1* {y 1 } , s *2 {y 1 , y 2 } ,…,
s*n {y1, y2 , y3..........yn } and let 1 ,2,…,n be the probabilities of selection of y1 , y 2 ,
…, y n respectively. Now define,
y1 Xi
d1 where α1
Nα1 X
1 y2 U2X2
d2 y
1 where α2
N α2 UiXi
i s1*
iΩ
Similarly,
1 yn Un Xn
dn y1 y2 y3 ......... where αn
N αn Ui Xi
i s *n -1
iΩ
1 n
T1 di
n i 1
… (3.2)
The form of estimator T1 is similar to the ordered estimator given by Des Raj (1956 a, b). On
similar lines, the estimate of variance of the estimator T1 which is also an ordered estimator
can be expressed as:
1 n
V̂(T1 )
n(n 1) i 1
(di T1 )2 ... (3.3)
The expression is similar to that of Des Raj’s estimate of variance of the ordered estimator of
population mean.
3.4.2 Stratified Contiguous Unit Based Spatial Sampling (Stratified CUBSS)
Stratified sampling consists in classifying the population units into a number of strata and
then selecting samples independently from each stratum. An appropriate estimator for the
population is obtained by suitably combining the stratum wise estimators of the character
under study. Stratification, apart from increasing the precision of the estimated population
IV.277
mean is also commonly used to provide estimates for the different sub-divisions constituting
the population. Here in this study, homogeneous zones formed on the basis of spatial
correlation can be considered as strata. Also strata can be formed by geographical boundaries
such as districts in the state or Tehsils in the district. The sample is selected from each
stratum using CUBSS technique, described in the previous section.
Let T2 be the estimator of the population mean obtained by applying stratified CUBSS, then
T2 is given by,
L
T2 Wh y h … (3.4)
h 1
nh d hj
where, y h is the sample mean of hth stratum and
j1 nh
1 y hn h U hj X hj
d hj y h1 y h2 y h(n h 1) where α hn h
N α hn h U hj X hj
jS*h(n h 1)
Since, the sample in the hth stratum is selected using the CUBSS technique described in
previous section, the variance of the estimator T2 can be given as,
L 1 nh 2
V̂(T2 ) w 2h (d y ) … (3.5)
n h (nh 1) j1
hj h
h 1
4. Case Study
The study has been carried out for the Rohtak district of Haryana to estimate irrigated area of
the district. Village wise data from District Handbook of Census (DHC), for the year 1991
has been used for the study. The village boundary map was digitized for 492 villages using
PC-ARC/INFO, a GIS software. The irrigated area of the village is treated as the character
under study (Y) and total cultivated area is the auxiliary character (X) for the study. One
thousand samples of different sizes have been selected following different sampling
procedures like Simple Random Sampling with Replacement, Simple Random Sampling
Without Replacement, Stratified SRSWOR, Dependent Unit Sequential Technique (Arbia,
1993), Stratified DUST and the proposed sampling technique.
4.1 Data Preparation
Two types of data are required for this study: spatial data and attribute (non-spatial) data.
The spatial and attribute data are prepared separately and then the two datasets are merged.
4.1.1 Spatial Data
Rohtak district consists of five tehsils namely, Maham, Bahadurgarh, Rohtak, Jajjar and
Gohana. Maps of these five tehsils were procured from District Handbook of Census (DHC),
IV.278
1991. The village boundaries of these five tehsil maps were digitized. The five digitized tehsil
maps were then merged to obtain the entire village wise map of the district.
4.1.2 Digitization of Maps
Digitization is the process of converting the map in the digital coordinates. Digitization of
each tehsil map has been done using SUMRGRID V digitizer, which has been configured
with GIS software PC ARC/INFO. The entire process has been carried out in ARCEDIT
module of the software. Digitization has involves following important steps.
i. Preparation of map sheets for digitizing.
ii. Digitization of the coverage map.
iii. Identification and correction of digitizing errors.
iv. Defining features and building topology.
v. Identification and correction of topology errors
The map obtained through digitization in ARCINFO is termed as coverage. The feature
topologies in this study are arcs and polygons. An area enclosed by arcs is called polygon.
Each village in the map refers to a polygon in the coverage. The digitized maps of five tehsils
were transformed to a common scale and were then merged to obtain a single (vector) map of
Rohtak district containing 492 villages. Hence, the coverage of Rohtak district consists of
492 villages i.e. 492 polygons. Now onwards villages will be referred to as polygons
throughout the study. Each polygon carries a unique identification number called ID number,
which is used to link attribute data with the spatial data. The data files such as Arc Attribute
Table (AAT) and Polygon Attribute Table (PAT) were created along with the map layer in
the process of digitization. The AAT and PAT are database format (DBF) files. AAT file
contains the spatial data and PAT file contains the information about this spatial data. These
files have been used further for linking spatial data and attribute data, for identification of
neighbours, calculation of spatial correlation sample selection and estimation process.
4.1.3 Attribute Data (non-spatial data)
The census data contain information on various important parameters of the village. In order
to classify the most important auxiliary character for improving the estimation of the main
character under study, the irrigated area, seven variables namely area under cultivation,
irrigated area, population of the village, number of households, area not available for
cultivation, culturable waste and wasteland were considered. The correlation matrix for the
above mentioned seven variables was computed. It was found that the cultivated area has the
highest correlation (r=0.9) with the study character among all other variables. Thus,
cultivated area was chosen to be the most appropriate auxiliary variable for the study. The
data at village level for these variables has been obtained from DHC of Rohtak for the year
1991. The data for the character under study, the irrigated and the auxiliary character, the
cultivated area, for each village is stored in a separate DBF file.
4.1.4 Linking Spatial and Attribute Data
The attribute data for irrigated area and cultivated area of each village has been linked with
the spatial data through PAT file using TABLE module of PC ARC/INFO. The PAT file is a
DBF file. Hence, the file containing attribute data is also in database format (DBF). The
newly created file containing the data for irrigated area and cultivated area for each village is
linked with the PAT file. One of the fields in both the files is the ID number of each polygon,
which is unique and is used for linking the two files.
IV.279
4.2 Identification of Neighbours
The contiguity based neighbours were identified in case of regular area units for each
polygon following Queen’s method which implies that all the polygons adjacent to a polygon
(x) i.e. either touching corners or touching the edges are identified as 1st lag neighbours of x.
Further, all the polygons touching the boundaries, both the corner and edges of the 1st lag
neighbours of x are identified as 2nd lag neighbours of the same polygon. Similarly,
neighbours of higher order lags have been identified. An algorithm was developed in
FORTRAN 90 to identify the contiguity neighbours using AAT file. The total number of
neighbours and the neighbours identified for each polygon has been stored in a separate file,
which is further used for calculating the spatial correlation.
The distance based neighbours were identified in case of irregular area units.The optimum
distance calculated on the basis of maximum correlation came out to be 20km. The criterion
used for identification of distance based neighbours at various spatial lags is:
i. If the distance between the centroids of any two polygons is less than or equal to 20
km then the two polygons are regarded as first lag neighbours.
ii. If the distance between the centroids of any two polygons is between 20 km to 40 km
then the two polygons are regarded as second lag neighbours.
iii. Similarly if the distance between the centroids of any two polygons is between 40 km
to 60 km then the two polygons are regarded as third lag neighbours and so on.
4.3 Calculation of Spatial Correlation and Testing of its Stationarity
An algorithm has been developed in FORTRAN 90 to calculate the spatial correlation for the
auxiliary variable. The PAT file and the file containing the neighbours have been used as the
input files for the algorithm. The value of the spatial correlation for each polygon is stored in
a separate file. The spatial correlation has been computed using expression given by Moran
(1950). Once, the spatial correlation for each polygon is computed, next step is the formation
of homogenous zones on the basis of spatial correlation. The zones are formed such that
spatial correlation of all the polygons falling in that zone is stationary. The PAT file and the
file containing the spatial correlation of each village have been used for making the zones.
The spatial correlation of the villages did not vary much therefore the entire district came as a
single zone on the basis of spatial correlation.
4. 4 Method of Sample Selection
Sample is selected according to the method mentioned in section 3.3 for both the cases of
regular and irregular area unit case.
4.5 Estimators used in the Study
Besides the proposed estimators mentioned in section 3.4 some other estimators of existing
sampling techniques, have also been considered for comparing the efficiency of the proposed
estimators. These estimators are for (i) Simple Random Sampling with Replacement (ii)
Simple Random Sampling Without Replacement (iii) Stratified SRSWOR (iv) Dependent
Unit Sequential Technique (DUST) (Arbia, 1993) and (v) Stratified DUST.
5. Conclusions
The results of the study shows that in spatial surveys considerable gain in efficiency could be
achieved by using a GIS assisted sampling strategy. It has been observed that spatial surveys
can be substantially improved by employing computer-intensive techniques that can easily be
handled by the modern power of GIS. In fact through Stratified CUBSS a better allocation of
IV.280
resources can be achieved leading to higher levels of accuracy of the estimates for a given
sample size or alternatively at a smaller sample size i.e. smaller cost for the same level of
accuracy.
References
1. Arbia, G. (1990). Sampling dependent spatial units. Workshop on Spatial statistics.
Commission on mathematical modeling of the ICU, Boston.
2. Arbia, G. (1991). GIS–based sampling design procedures. Proc. of EGIS, 1991, 1,
EGIS foundation Utrecht.
3. Arbia, G. (1993). The use of GIS in spatial statistical surveys. Int. Stat. Rev., 61(2),
339-359.
4. Brunsdon, C., Fotheringham, S. and Martin, C. (1998). Geographically weighted
regression modeling spatial non-stationarity. The Statistician, 47 (3), 431-443.
5. Cliff, A.D. and Ord, J.K. (1973). Spatial autocorrelation. Pion, London.
6. Cliff, A.D. and Ord, J.K. (1981). Spatial Processes. Pion, London.
7. Cochran, W.G. (1963). Sampling Techniques. Wiley.
8. Hedayat, A.S., Rao, C.R. and Stufken, J. (1988a). Sampling plans excluding
contiguous units. J. Stats. Plan. Inf., 19, 159-170.
9. Hedayat, A.S., Rao, C.R. and Stufken, J. (1988b). Designs in Survey sampling
avoiding Contiguous Units. Handbook of Statistics P.R. Krishnaiah and C.R. Rao.
Eds. 6. Elsevier Science Publishers B.V. 575-583.
10. Kotz, S. and Johnson, N.L. (1986). Spatial Sampling. Encyclopedia of Statistical
Science. 6 Wiley Intersc. Pub. John Wiely and sons.
11. Moran, P.A.P. (1950). Notes on continuous stochastic phenomena. Biometrika, 37,
17.
12. Raj, D. (1956a). Some estimators in sampling with varying probabilities without
replacement. J. Amer. Stat. Assoc., 51, 269-284.
13. Raj, D. (1956b). A note on the determination of optimum probabilities in sampling
without replacement. Sankhya,17, 197-200.
14. Ripley, B.D. (1981). Spatial Statistics. Wiley, New York.
15. Sahoo, P.M., Singh, R. and Rai, A. (2006) Spatial Sampling procedures for
Agricultural Surveys using Geographic Information System. Journal of Indian Society
Agricultural Statistics. 60 (2), 134-143.
16. Unwin, A. and Unwin, D. (1998). Exploratory spatial data analysis with local
statistics. The Statistician, 47, 415-421.
IV.281
SPATIAL ANALYTIC HIERARCHY PROCESS FOR
IDENTITYFICATION OF POTENTIAL AGROFORESTRY
AREAS USING GIS
Tauqueer Ahmad
I.A.S.R.I., New Delhi -110012
1. Introduction
Producing food in adequate quantities and of good quality for the fast growing population is
one of the most important problems faced by the developing countries. The problem becomes
more difficult, because it is necessary not only to devise means of producing this basic need,
but also to ensure that the quality of land resources that are directly or indirectly utilized in
producing food is maintained and improved. Therefore, it is important to increase
productivity and at the same time to conserve and enhance the quality of eco-system.
Forests are one of the most important renewable natural resources that contribute
substantially to the social and economic development and also play a major role to enhance
the quality of environment. Since it is well known that India does not possess many degrees
of freedom with respect to utilization of new arable land for food production, so it becomes
imperative to devise land management and farming systems that are not only capable of
producing food from marginal agricultural land but also capable of maintaining and
improving the quality of producing environment. One such system is agroforestry that helps
in providing all necessities of life and at the same time maintaining the quality of land
resources. The integration of farming with forestry practices on the farm to the benefit of
agriculture is known as agroforesty.
It is therefore, necessary to consider the factors responsible for growth of agroforestry while
examining the relevance of agroforestry systems to social, economic and environmental
development. Agroforestry plays a vital role in achieving integrated rural and urban
development that is of great importance for planners. The most important requirement for
recommending/adopting agroforestry system is identification of potential areas for growth of
agroforestry.
Keeping in view the above facts, a study entitled “Development of GIS based technique for
identification of potential agroforestry areas” was undertaken in Yamunanagar district of
Haryana State. Prior to this study, a pilot study was conducted by IASRI in Chhachhrauli
block of Yamunanagar district to see the impact of agroforestry on socio-economic
conditions of the farmers. In this pilot study, exhaustive information related to agroforestry
was available. This study was undertaken in the Yamunanagar district utilizing the primary
data available in the pilot study and using Geographic Information System (GIS), the recent
information technology tool for identification of potential agroforestry areas.
The main objective of the study was to develop a GIS based technique for identification of
potential agroforestry areas. The first step in identification of potential agroforestry areas was
to identify the important factors responsible for growth of agroforestry.
2. Identification of Important Factors in the District
A soft copy of 1991 census data of Yamunanagar district was obtained from Data Centre,
office of Registrar General, Pushpa Bhawan, Madangir, New Delhi. The data on area under
agroforesty for 20 villages of the district was available from earlier survey (pilot study) which
was compared with the 1991 census data based on certain variables like area of the village,
total population of the village and number of households in the village. Once the 20 villages
were identified in the census data, area under agroforestry for these villages was taken from
earlier survey while rest of the variables (54 in number) were taken from the census data.
This database was finally used for the purpose of analysis. First of all correlation matrix was
obtained using SPSS package. Then using SPSS package, regression analysis techniques such
as Stepwise Forward and Backward regression techniques were used for identifying the
important factors responsible for growth of agroforestry. ‘Area under agroforestry’ was
treated as dependent variable while other factors highly correlated with ‘area under
agroforestry’ were taken as independent variables. On the basis of above mentioned
regression analysis techniques, the following important factors responsible for growth of
agroforestry were identified:
1. Male agricultural laborers
2. Total irrigated area
3. Total cultivated area
4. Total number of literates
5. Male cultivators
3. GIS based Objective Spatial Analytic Hierarchy Process for Identification of
Potential Agroforestry Areas
This study strives to find out the potential agroforestry areas using GIS and proposed
Objective Analytic Hierarchy Process (OAHP). The resulting technique named Objective
Spatial AHP ranks the areas based on area attributes and depicts the potential agroforestry
areas in the map. In this study, villages were treated as ‘areas’ and potential agroforestry
villages of the district were identified using the proposed objective Spatial AHP.
The proposed Objective Spatial AHP is a method that combines GIS and proposed OAHP to
identify and rank areas that were suitable for agroforestry, using the statistical techniques
involved in the proposed OAHP and data contained in GIS maps. The proposed Objective
Spatial AHP involves the following five steps:
1. Identifying the issues and objectives (i.e. the decision factors) associated with the
problem
2. Structuring them in a decision-hierarchy
3. Determining the relative importance of decision-hierarchy elements
4. Aggregating these measures in order to develop a suitability index of alternatives
5. Ranking the spatial objects (polygons or raster) according to the suitability index.
3.1 Identification of Decision Factors and Decision Hierarchy
Decision factors (level 2) were identified using the regression analysis technique, treating
goal (level 1) as dependent variable and all other variables that were highly correlated with
dependent variable as independent variables.
The decision factors under this study were nothing but the identified important factors
responsible for growth of agroforestry explained in section 2.
Once the decision factors for a given problem were identified, they were arranged in a
decision hierarchy. A decision hierarchy consisted of a number of levels. The top level was
the goal of the hierarchy, while the rest of the levels were sub factors and sub-sub factors etc.
The sub factors (level 3) and subsequent levels were identified using regression analysis
IV.283
technique treating decision factors (level 2) and sub factors (level 3) respectively as
dependent variables while all other variables correlated with decision factors and sub factors
respectively as independent variables. The lowest level (cell attribute values) was obtained
based on the percentile values of the next higher level.
In this study, goal was to identify areas most suitable for agroforestry. Therefore, the top
level i.e., level 1, of the decision hierarchy was ‘area under agroforestry’.
The major decision factors, level 2, were male agricultural laborers, total irrigated area, total
cultivated area, total number of literates and male cultivators.
Level 3 represented sub factors. Sub factors were obtained using regression analysis
technique treating decision factors as dependent variables and the variables obtained from the
1991 census data of Yamunanagar district that were correlated with the corresponding
dependent variable as independent variables.
Level 4 represented the various values that could be assigned to each alternative. Thus, each
alternative was described by the combination of level 4 values assigned to it. Factors directly
connected to the lowest level were called here “attributes”. For example, “village population”
under “male agricultural laborers” was an attribute. The lowest level of the decision hierarchy
consisted of the attribute values assigned to specific alternatives based on the percentile
values of the sub factors. The 25th percentile had been categorized as low, 25th to 75th
percentile as medium and 75th to 100th percentile as high. So there were three sub-sub factors
as low, medium and high for each sub factor under this study. Thus, lowest level described
each alternative.
The decision hierarchy used in this study used data that were available from the 1991 census
and from the previous study.
In the proposed Objective Spatial AHP, the objective was to identify and rank areas. To
accomplish this objective, the study area was divided into cells, and proposed OAHP was
used to rank each cell. In this study, the district consisted of 654 villages. Each village had
been considered as a cell.
The suitability of each village i.e. cell can be considered to be a function of the utility of
cell’s attribute values. In Objective AHP, the cell attribute values were grouped into finite
numbers of categories, based on the pre-existing form of data and percentile values.
Examples of cell attribute categories shown in figure 1 included low, medium and high for all
the sub factors.
IV.284
Level 1 Level 2 Level 3 Level 4
Goal Decision factors Sub factors Sub-sub factors
(Cell attribute values)
Low Medium High
Low (<72.77%
Cultivated (A1) Medium (72.77-89.83%)
Area High (>89.83%)
Male Agricultural Low (<358)
Laborers Village (A2) Medium (358-1117)
(A) Population High (>1117)
Low (<50%)
Area Irrigated (B1) Medium (50-90%)
by tube wells High (>90%)
Low (<72.77%)
Total Irrigated Cultivated (B2) Medium(72.77-89.83%)
Area Area High (>89.83%)
(B) Low (<0.068%)
Geographical (B3) Medium(0.068-0.164%)
Area High (>0.164%)
Low (<0.068%)
Geographical (C1) Medium(0.068-0.164%)
Area under Area High (>0.164%)
agroforestry Total cultivated Low (<52)
Area Number of (C2) Medium (52-165)
(C) Households High (>165)
Low (= 0)
No. of Primary (D1) Medium (= 1)
Schools High (>1)
Low (= 0)
Total No. of No. of Sec. (D2) Medium (= 1)
Literates Schools High (>1)
(D) Low (<358)
Village (D3) Medium (358-1117)
Population High (>1117)
Low (<31.72%)
Number of (E1) Medium(31.72-49.72%)
Literates High (>49.72%)
Low (<37.58%)
Male Cultivators Total Irrigated (E2) Medium(37.58-87.87%)
(E) Area High (>87.87%)
Low (72.77%)
Cultivated (E3) Medium(72.77-89.83%)
Area High (>89.83%)
IV.285
3.2 Relative Importance Weights for Different Factors
Pair wise comparisons of all related attribute values were used to establish the relative
importance of hierarchy elements. Pair wise comparisons are generally based on the decision
matrix obtained as per decision makers or experts’ judgment. In the proposed procedure,
decision makers or experts have no role to play in finding out the elements of decision
matrix. The elements of decision matrix were obtained as a ratio of R 2 of the two attributes
that are pair wise compared. The lowest level attributes were pair wise compared using the
numerical scores given based on statistical background.
The diagonal of the decision matrix always contains ones because it contains comparisons of
identical attributes. The cells above and below diagonal are reciprocals.
The relative importance weight (RIW) assigned to each hierarchy element was determined by
normalizing the eigenvector of the decision matrix. Eigenvector values were estimated by
multiplying all the elements in a row and taking the Nth root of the product, where N was the
number of row elements (Saaty 1980). Normalization of the eigenvector was accomplished
by dividing each eigenvector element by the sum of the eigenvector elements. The modified
RIWs were obtained by multiplying the RIWs with the corresponding cell values, which were
the values of the variables obtained after suitable transformation.
Regression analysis was done treating area under agroforestry as dependent variable and
decision factor 1 as independent variable using the data of 20 completely enumerated villages
and corresponding census data. The same was done for all the decision factors.
The decision matrix for pair wise comparison of the decision factors is given by:
a 11 a 12 a 13 a 14 a 15
a a 22 a 23 a 24 a 25
21
D= a 31 a 32 a 33 a 34 a 35
a 41 a 42 a 43 a 44 a 45
a 51 a 52 a 53 a 54 a 55
w1 / w1 w1 / w 2 w1 / w 3 w1 / w 4 w1 / w 5
w / w w / w w 2 / w3 w2 / w4 w 2 / w 5
2 1 2 2
= w 3 / w1 w 3 / w 2 w3 / w3 w3 / w 4 w3 / w5
w 4 / w1 w 4 / w 2 w 4 / w3 w4 / w4 w 4 / w5
w 5 / w1 w 5 / w 2 w5 / w3 w5 / w 4 w 5 / w 5
IV.286
Hence, the decision matrix and the relative importance weights of the decision factors are
given below in Table 1:
Estimated
A B C D E eigen RIW
element
A 1.0000 1.0191 1.0299 1.2665 1.7987 1.1904 0.2326
B 0.9812 1.0000 1.0106 1.2428 1.7650 1.1681 0.2283
C 0.9709 0.9895 1.0000 1.2297 1.7464 1.1558 0.2259
D 0.7895 0.8046 0.8131 1.0000 1.4201 0.9399 0.1837
E 0.5559 0.5665 0.5725 0.7041 1.0000 0.6618 0.1293
Total 5.1163
Similarly, relative importance weights of all the sub factors were obtained using the census
data of all 654 villages and applying regression analysis technique treating one of the
decision factors as dependent variable and corresponding sub factors as independent
variables.
Sub-sub factors for each sub factors were (1) High, (2) Medium and (3) Low. ‘High’ was
given score 9 and ‘Low’ was given score 1 following the conventional numerical scales. The
average of scores 9 and 1 is 5 and hence ‘Medium’ was given score 5. Therefore, the values
for pair wise comparisons are as follows:
Preference of ‘High’ (9) with ‘Low’ (1) = 9-1+1 = 9
Preference of ‘High‘(9) with ‘High’ (9) = 9-9+1 = 1
Preference of ‘High’ (9) with ‘Medium’ (5) = 9-5+1 = 5
Preference of ‘Medium’ (5) with ‘Low’ (1) = 5-1+1 = 5
Hence, the decision matrix of the above mentioned sub-sub factors and corresponding
relative importance weights are given below in Table 2:
IV.287
the weights for all the villages. Therefore, a method to find out the revised relative
importance weight (RRIW) was proposed under this study.
The modified RIWs for each cell were obtained by multiplying the RIWs with the
corresponding cell values that were the values of the variables obtained after suitable
transformation.
Finally, the revised relative importance weight (RRIW) of each factor of a particular level for
each cell was obtained as a ratio of the modified RIW of that factor and sum of the modified
RIWs of all the factors of that level.
3.4 Suitability Index and Suitability Rank
A suitability index for each raster cell was determined by aggregating revised relative
importance weights (RRIWs) at each level of hierarchy. Suitability indices were calculated
by multiplying the RRIWs of a cell’s attribute values by the RRIWs of the associated higher
level factors, summing the values for all grouped elements, multiplying those sums by the
RRIWs of the associated higher level factor, and repeating until level 2 is reached. The
equation for a four level hierarchy is:
N2 N3i
SI = RRIW 2 3
i RRIWij RRIWijk
4
... (1)
i j
where SI = Suitability Index; N2 = the number of level 2 decision factors; RRIWi2 = the
revised relative importance weight of level 2 decision factor i; N3i = the number of level 3 sub
factors directly connected to level 2 decision factor i; RRIWij3 = the revised relative
importance weight of level 3 sub factor j of level 2 decision factor i; and RRIWijk4 = the
revised relative importance weight of level 4 attribute category k of level 3 sub factor j and
level 2 decision factor i (If a decision hierarchy has more or fewer levels, the formula must be
modified appropriately).
The suitability index of each raster cell (village) was obtained using the above-mentioned
formula where RRIWs were the revised relative importance weights (village wise) based on
the modified RIWs.
After determining the suitability index for each village, percentile values were obtained and
villages with values less than or equal to 25th percentile were grouped and identified as the
low potential villages for agroforestry. The villages with values greater than 25th percentile
and less than or equal to 75th percentile were grouped as moderate potential villages for
agroforestry and the villages having values greater than 75th percentile were grouped as high
potential villages i.e. the most suitable villages for agroforestry. The 25th percentile of
suitability indices was 0.138345 whereas 75th percentile was 0.322428.
All the villages were categorized as (1) High potential villages, (2) Moderate potential
villages and (3) Least potential villages based on the suitability rank.
Ranking of each raster cell according to suitability index and mapping of areas potential for
agroforestry require attachment of suitability indices for all 654 villages (raster cells) with the
digitized district map.
The following steps are involved to reach to the mapping of potential agroforestry areas:
IV.288
(i) Digitization of the District Map with Village Boundaries
The district map with village boundaries was not available. Yamunanagar district consists of
two Tehsils Chhachhrauli and Jagadhari. Tehsil maps with village boundaries were available
in the Yamunanagar District Hand Book of Census 1991. All the villages of Chhachhrauli
and Jagadhari tehsils were digitized separately using Tehsil maps with the help of ARC Info,
a GIS software, and digitizer available in the Remote Sensing Lab of IASRI, New Delhi.
There were approximately 1800 arcs after digitization, which were edited using the same GIS
software. Both the Tehsil maps were at different scale and were to be joined together. For
joining the maps, georeferencing was must.
(ii) Georeferencing of the Digitized Maps
Georeferencing of the maps require toposheets to know the longitudes and latitudes of the
maps. The longitudes and latitudes of the most of the boundary points for both the Tehsil
maps were taken from the toposheets of Yamunanagar district obtained from the Survey of
India office, Janpath, N. Delhi. Using the latitudes and longitudes, both the Tehsil maps were
georeferenced and brought into the same scale. Both the georeferenced Tehsil maps were
joined together and hence the digitized district map with village boundaries was obtained.
(iii) Attachment of Attribute Data with the Map
All the villages in the map have their location numbers. These location numbers in the map
were the last three digits of the 18-digited village location code of the 1991 census data. A
unique location code (5 digited) comprising Tehsil code (two digited) and village code (three
digited) of the census data was created corresponding to each polygon (village) of the map.
The 18-digited village location code was trimmed to the 5-digited code (comprising Tehsil
code and village code) created as above. Then attribute data including the suitability index
were attached to the map with the help of the unique field i.e. village location code. Now by
clicking on any village on the map, information like suitability index, percentage literate,
percentage irrigated area, percentage cultivated area, village population, geographical area,
number of households etc. of that particular village can be seen.
(iv) Identification of Potential Agroforesty Areas
Giving the criteria of suitability index greater than 75th percentile had generated a map of
potential villages for agroforestry in Yamunanagar district. In the same map, moderately
suitable villages for agroforestry and least suitable villages for agroforestry were depicted
using the criteria of suitability index between 25th percentile and less than or equal to 75th
percentile and the criteria of suitability index less than or equal to 25th percentile respectively.
Hence, all the villages of Yamunanagar district categorized as (1) High potential villages, (2)
Moderate potential villages, and (3) Least potential villages based on the suitability indices
were depicted in three different colours in the same map named as “Agroforestry suitability
index map of Yamunanagar district”. Clicking on any particular village of the district on this
map can show the attributes of that village.
References
1. King, K.F.S. (1981). Some principal of agroforestry-Key note address in the
“Proceedings of the agroforestry seminar “ held at Imphal on May 16-18, 1979. Eds:
Jaiswal, P.L.
2. Rai, A., Srivastava, A.K., Singh, M., Singh, S. and Sapra, R.K. (1999). A pilot study
of agroforestry in Chhachhrauli block of Yamunanagar district (Haryana).Project
Report, IASRI, New Delhi.
IV.289
3. Saaty, T.L. (1979). The US-OPEC energy conflict-The payoff matrix by analytic
hierarchy process. Int. J. of Game Theory, 8(4), 225-234.
4. Saaty, T.L. (1980). The analytic hierarchy process. McGraw-Hill, Inc., New York,
N.Y.
5. Siddiqui, M.Z., Everett, J.W., and Vieux, B.E. (1996). Landfill siting using
Geographic Information System: A demonstration. Journal of Environmental
Engineering, 122(6), 515-523.
6. Singh, B. (1981). Scope of agroforestry in Panjab and Himachal Pradesh. Proc. of the
agroforestry seminar held at Imphal on May 16-18, Eds: Jaiswal, P.L., pp.147.
7. Singh, G.B. and Pazo, P.O. (1981). Agroforestry in Eastern Himalayas. Proc. of the
agroforesry seminar held at Imphal on May 16-18, Eds: Jaiswal, P.L., pp. 27.
8. Toky, O.P.and R.P. Bisht. (1992). Observation on rooting patterns of some
agroforestry trees in arid regions of northwestern India. Agroforestry Systems, 18,
245-264.
IV.290